GeekChronicles Issue 6 – GeekyAnts' AI Era, Product Wins & Community Buzz

Scan for Digital Copy GEIK CHRO NiCLES by © GeskyAnts PREFACE This publication draws from inside the work. It follows projects across design, engineering, operations, and editorial practice. The entries are drawn from transcripts, case studies, feature notes, and internal updates—each reflecting decisions already made and systems already in use. It includes notes from internal townhalls, platform updates, design reflections, and DevOps workflows. Some pieces document changes in process or structure. Others focus on specific events, releases, or production timelines. Subjects covered range from rollout coordination and interface design to legal policy, branding, and team retrospectives. All of the pieces together give a holistic image of how different parts of the organisation are planning, building, releasing, and documenting their work across projects and functions. The issue follows a continuous line of activity, with each entry contributing to an ongoing record of practice. The Minds Behind the Magazine This Issue Wouldn't Exist Without The Dedication And Support Of The Incredible Individuals Who Helped Shape It. We're Grateful For Their Contributions And Belief In Our Mission To Inform, Inspire, And Connect. Annalisa J Urumbath Senior Account Manager Rajat Chaudhary Software Engineer II Takasi Sandeep Tech Lead | Christine Mathias Manager Legal Vineeth Kiran UI / UX Designer Jyotsna Chetan Hegde Account Manager UI/UX Design Lead II Robin Mathew UI / UX Designer ( al ii oianket Sahu a an | Raksha S nnovation ead DevOps Engineer j Ul UX Designer Paridhi Tulsian Digital Marketing Lead Mithun K : DevOps Engineer Sarthak S Bakre Sachin M Software Engineer II UI/UX Designer | Joydeep Nath Tech Lead I! Core Contributors Editorial Team Designer Content Editing Content Curation Aswathy Anil Neeraj Yadav Amrit Saluja Prem Goswami Rakesh Ningthoujam Boudhayan Ghosh Prince Thakur © 2025 GeekyAnts. All rights reserved. No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the publisher. All content, including text, images, and design, is the intellectual property of GeekyAnts. For permissions or inquiries, please contact magazine@geekyants.com TABLE O F CONTENTS Editor’s Notes & Value of the Month The GeekyAnts Take WWDC 2025: Liquid Glass Apple shifts UI to fluid interaction Claude 4 Review Smarter Al for developers Shaper.studio Designers explore code-as-design IOS 26 UX Shift Apple makes UX predictive Build vs Partner In-house isn't always faster Al Estimation Tool Smarter project scoping with Al Legal Update Key policy and visa updates React Native Townhall Real RN solutions from the field 29 Design Sprint Recap Fast delivery through structure Agentic Al in Fintech Al agents now handle workflows US Al Meetup Agents, builders, and new roles36 Fintech Case Study Scalable global payment app39 Al Townhall Insights Practical Al deployment stories 49 45 AM Team Wins Clients, feedback, and new leads DevOps Meetup Talks of tools to platforms Al + Design Al assists, humans decide GeekWiz Awards Celebrating standout performers The Overthink Tank Bold thoughts on design and tech. Editor’s Notes For years, a “good product” meant immaculate design: clean visuals, intuitive user experience, and a polished user journey. If you added a subtle moment of delight and tight performance, you could count on applause and adoption. That handbook is now obsolete. ChatGPT reset the rules. A single prompt can deliver answers faster than months of traditional design sprints. When an artificial intelligence model solves a user's problem before the question is fully typed, screen aesthetics and carefully nested menus become irrelevant. The result is sometimes raw and awkward, yet it is ten times quicker, more flexible, and laser- personalised. Experience has given way to outcome. In this landscape, intelligence is the new gold. Capability, contextual awareness, and self- learning behaviour now define product quality. Users no longer care about navigating your application; they care about immediate results. Software that cannot adapt, remember, and reason is instantly disposable. Static workflows are already obsolete, and grafting a large language model onto legacy code is merely makeup on a corpse. The bar is now zero-friction interaction and systems that quietly eliminate onboarding, documentation, and training because they already understand what each user needs. Microsoft saw the future first. Copilot, GitHub’s artificial intelligence pair-programmer, and language model-infused Office experiments signalled bold intent. Yet the company wrapped these breakthroughs in enterprise bundles and compliance slides, muting the narrative. They possessed the right ingredients but treated intelligence like an add-on rather than the headline. OpenAl seized that spotlight, making advanced capability impossible to ignore. Apple, characteristically late but precise, is turning heads with Liquid Glass. More than an augmented display, it signals a world where intelligence is the interface. Tap, click, and swipe give way to glance, voice, and intent. As Apple often does, it is not inventing the technology so much as defining its language, presenting a future where the boundary between user and system dissolves into glass. The message is simple: the user interface is no longer the product; the thinking engine beneath it is. The industry now values products whose core loop is adaptive reasoning, whose interfaces vanish when unnecessary, and whose builders act more like systems designers than visual artists. Teams that embed memory, context, and evolution into version one will lead. Beautiful dashboards and delightful onboarding belong to the pre-ChatGPT world; today’s expectation is a working prototype that feels alive from day one. If your software cannot learn and anticipate, it is already irrelevant, and no pitch deck—no matter how polished— will bridge that gap. O1 Future Ready /'fjuz.tfa 'red.i/ How do we become future-ready if one is not a prophet? Being future-ready is deeply rooted in being ready for the S/ present. Only by deeply knowing today can we prepare meaningfully for tomorrow. At Geekyants, we don’t dwell on the past—we look ahead by fully embracing the present: what it holds, what it offers, and what it teaches. The quality of being progressive is a virtue, isn’t it? Letting go of the bygones, making space for what wants to emerge. A future that values creativity, structure, and imagination in equal measure. Valuing invention over routine, questioning assumptions that no longer serve the work. Let us stay present. Let us stay open. The future will follow. 03 DESIGN PODCAST ° Through the Glass: Three Designers Reflect on WWDC 2025 pple’s WWDC 2025 keynote unfolded like a controlled study in visual experimentation. The presentation offered no sweeping declarations. It instead introduced a series of deliberate shifts—some surface-level, others structural. At the centre was a visual system called Liquid Glass, a new layer of interaction built on refraction, elasticity, and fluid movement. Vineeth Kiran, Raksha S, and Robin Mathew, UI/UX designers at GeekyAnts, watched the event closely and recorded their reactions. What emerged was not a rundown of features, but a layered reading of material, interface, and direction—what had changed, what felt deliberate, and what resisted understanding. X » . = Backseat Driver Kane Brown | gg = TIMESENSITIVE . | Grab groceries fo ” Today, 9:25AM ry — Liquid Glass and the Edges of Material Design The Liquid Glass interface was the dominant visual signature of the keynote. It was animated with tension. Ul blocks swelled and contracted as if suspended in fluid. Text refracted as elements moved beneath it. One designer described it as the feeling of water on glass—a visual behaviour more than a decorative filter. Another noted its structure: soft at the edges, rubber-like, almost elastic in its visual rhythm. None of them saw it as “glassmorphism.” That comparison was dismissed early. The distortion here was physical. It mimicked actual light bending and object displacement. The implementation showed clear effort in microphysics, not aesthetic layering. What concerned them more was not what it looked like, but whether it could function across unpredictable environments. Mixed backgrounds caused problems. When Liquid Glass passed over content with mid-tone contrast or layered video, text clarity fell apart. The system had rules, but many real-world contexts ignored those rules. The visual field became unstable. The underlying principle—dynamic adaptation— was understood. Foreground elements adjusted their tone based on what lay behind them. However, this system, in practice, had gaps. Accessibility suffered in edge cases, and no solution had been presented for those. Performance came next. Animating distortion across layers increased GPU load. The fluidity came at a cost. While keynote demos showed precision, these designers questioned how reliably this would translate across the iOS ecosystem. Designing What Cannot Be Designed There was an immediate technical frustration: this cannot be prototyped in Figma. The team discussed how none of today’s standard design tools supported real-time distortion, interactive light refraction, or layered visual liquidity. There was no way to replicate what Apple had built. That absence was more than technical. It exposed a directional shift. Design tooling had, until now, evolved in parallel with platform behaviour. With Liquid Glass, the platform had pulled ahead. Designers would be asked to build experiences they could not simulate. There was speculation that plugins might emerge. Perhaps Figma would adapt. But for now, this was an interface language inaccessible to those tasked with extending it. SS Q @BPe Teapri 9:41AMG(s +) * DESIGN PODCAST Space and Orientation VisionOS expands how interface elements behave in a physical context. Screens are no longer the only place where interactions occur. Components respond to spatial placement, align with surfaces, and adjust based on the environment. In one demonstration, the weather widget was positioned on a wall. Once placed, it revealed a view showing outdoor conditions. The action responded to the physical surface, not the screen. Placement shaped what the interface did, and the wall became part of the interaction. FaceTime introduced a different use of orientation. In addition to updated avatar rendering, the system allowed one user to shift into the other person’s point of view. This altered how presence was structured inside the conversation. Interaction was built around mutual perspective, not a fixed display. Together, these features show a direction where space is not visual context but functional input. Interfaces change depending on where they are, what they face, and how they relate to the environment around them. Track ‘s v ) 2 arth. Title vo o\{e) JJ CC) x 04 O05 DESIGN PODCAST ° Fragmented Intelligence The integration of ChatGPT into Apple's ecosystem brought new capabilities, but also surfaced a gap in interface logic. The system can now process screenshots, summarise content, and respond to natural language queries. These actions are handled through ChatGPT, not Siri. Siri still managed simple tasks—reminders, conversions, alarms—but had not been extended to the new functions. One designer pointed out that asking Siri for a recipe still opened a browser tab. Another noted that screenshot queries now bypass Siri entirely. The interface sent them straight to GPT. These moments exposed a deeper issue. There was no unified layer guiding how intelligence was surfaced. The system responded in parts, but there was no point of contact that represented the whole. The group returned several times due to the absence of a unified system agent. Siri had not absorbed the new capabilities, and ChatGPT operated as a separate layer. This left no clear entry point for interaction. Intelligence was distributed across features, but the OS did not present it as a system. That design decision shaped how the tools were understood—not as extensions of a single assistant, but as isolated responses to specific tasks. Developer Adjustment Several interface behaviours were redefined. Ul alignment, motion thresholds, and visual layering had been updated across the system. These changes introduced new defaults that previous applications were not designed to meet. Apps that had once matched the platform visually would now stand apart if left unchanged. The group understood this shift as a reset in how native interfaces were structured. The required updates did not involve new technical models, but they did require adaptation. Existing layouts, transitions, and input logic had to be re-evaluated against the updated system standard. The work ahead was procedural, but the implications were broad. What had previously been accepted as consistent would now need to be revised. Making the Incomplete Visible Toward the end of the discussion, the group focused on Liquid Glass. Several limitations were already visible. The material did not render reliably across background conditions. Interface tools could not replicate their effects. Performance expectations remained undefined. The system had been released without full implementation support. There was agreement that the timing was intentional. By making the system visible before it was fully developed, Apple had opened it to feedback. The focus shifted toward its construction—how the interface handled distortion, how it moved in layered space, how it interacted with depth and light. These elements were being examined in detail, even though teams could not yet use them in production. The release created a reference point. Designers were studying the system, discussing its constraints, and outlining what would need to change to make it usable. The attention arrived before the platform was ready to support it. © * DESIGN PODCAST Closing Note The discussion moved across features, systems, and design assumptions. Some updates were accepted immediately. Others were left open, noted for what they implied rather than what they resolved. The group focused less on what had been added and more on how the operating system now expected users—and designers— to think. The conversation did not end with a single opinion or conclusion. It outlined the kinds of decisions that would follow. SUBSCRIBED % ‘p| Amrit Saluja Tare «Technical Content Writer Impact on Developer Workflows Performance benchmarks often reveal more than promotional claims, and Claude 4 backs its promises with impressive numbers. Claude Opus 4 clinched first place in the critical coding benchmarks, scoring 72.5% on SWE-bench and 43.2% on Terminal-bench. It brings hybrid thinking modes, parallel tool execution, and local file memory, allowing teams to store and retrieve project progress across sessions. Through Claude's dedicated Code SDK, IDE integrations, and Developer Mode, Opus 4 enables developers to debug, refine, and streamline their workflows. Surprisingly, Claude Sonnet 4 edged ahead on SWE-bench at 72.7%, outperforming competitors like GPT-4.1, Gemini 2.5 Pro, and its predecessor, Claude 3.7. It targets agile development, rapid iteration cycles, and coding agents, combining parallel tool use with hybrid reasoning to minimise shortcuts in complex problem-solving. * CLAUDE SERIES 4 What does this mean for businesses? Opus 4 positions itself for enterprise-grade development at $15 per million input tokens. Sonnet 4, with a broader reach, offers strong coding capabilities at $3 per million input tokens, lowering the barrier for individual developers, startups, and teams exploring Al-driven workflows. Such benchmarks confirm something crucial — Claude 4 means business, especially when you pair it with IDE integrations like VS Code, JetBrains, Cursor, Replit, and GitHub Copilot. It manages multi-file real-time edits, including autonomous tasks like playing Pokémon Red through persistent memory management. Yuvraj Singh, Software Engineer at GeekyAnts who tested Claude 4 during React Native (bare) development, said: “In my opinion, it adapts to the codebase better now, concerning the design and styling patterns as well as to state management, without much narrative being provided. Also, it tends to explain the changes and suggestions better than before, | reckon." 08 ACCURACY 80 60 40 20 79.4% with parallel test- time compute 80.2% with parallel test- time compute Claude Sonnet 4 Claude Opus 4 v Claude Sonnet 4 w Claude Opus 4 70.3% with parallel test- 72.1% time compute 69.1% 72.5% 72.7% 63.2% 62.3% 54.6% Opus 4 Sonnet 4 Sonnet 3.7 OpenAl OpenAl OpenAl Gemini Codex-1 03 GPT-4.1 2.5 Pro Preview (05-06) * CLAUDE SERIES 4 The GeekyAnts Take — Where Does Claude 4 Truly Fit In? Claude 4 integrated smoothly into our React Over time, Claude 4 has settled into the rhythm Native workflow, revealing its capabilities of our workflow. It handles the slow friction of through quiet, deliberate action. It understood repetition, the small corrections that often the rhythm of component structures, the escape attention, and the edge cases that layering of style sheets, and the tension between emerge late in the build. It does not seek to design intention and implementation. In early change the way we work, but it shapes what we use, it aligned naturally with our conventions, can focus on. With fewer interruptions and fewer offering adjustments that respected existing course corrections, teams can look ahead to logic while quietly improving its clarity. There design more thoughtfully, to solve problems at was no need to recalibrate the environment scale, and to build with greater momentum. around it. Claude 4 adjusted to us. What sets this model apart is its ability to operate with limited instruction. It tolerates ambiguity, responds with relevance, and applies context with accuracy. Code suggestions are not only correct in function, but also coherent in style. When asked to resolve layout inconsistencies or simplify state handling, it responds with solutions that feel thought through. Developers find themselves giving fewer directives and receiving more insight. The model does not just complete lines of code—it engages with the intent behind them. 66 11 SHAPER.STUDIO ° When Designers Met Shaper: What We Learned from Our First Alpha Test Paridhi Tulsian Digital Marketing Lead hange is the only constant. But in a tool like Shaper, it is also a compass," said Suraj Ahmed, our CTO, just before we ran our first real-world test. On May 22, 2025, three designers — each comfortable with Figma and deeply familiar with modern design workflows — logged into Shaper for the first time. What followed was not a polished walkthrough. It was real, hands- on exploration. They clicked around, got stuck, asked questions, and gave unfiltered feedback- exactly what we hoped for. There was no script and no walkthrough. The goal was to observe how quickly someone could pick up Shaper on their own. Could designers approach it the way they approach the tools they already use? More importantly, did it help them work better and faster \ Shaper Create with Al assistance Here is some text . =e Edit className = Position v Tailwind-native styling Here is some text static 3 v Git-native from Day One Your project lives in a real Git repo, ready for Vv developers to scale. Container Layout v Extend or edit code seamlessly Build real Uls visually, without touching You can jump into the same codebase inside code Flex v oS Shaper or in their favorite IDE. Canvas-first editing that outputs clean SEEING: Tailwind code. Size 100vw wv 2835px al 20 Overflow 13 © Flex Properties Resources Legals Contact Us 7 Column Vv bf What Landed Well Let’s be honest: no alpha session is complete without at least one “oh wow” moment — and Shaper delivered a few. The token system impressed all three testers. Changing a style token instantly updated everything connected to it — no need to dig into nested layers or do repetitive cleanup. "It is like having a design system that’s fun to use," said Designer 2. And onboarding? Surprisingly smooth. All three testers said they were able to start designing quickly, without needing a walkthrough or reading documentation. "This felt easier to start with than | expected. | was editing real stuff right away," said Designer 1. The biggest “aha,” though, came when they realised they were not designing a mockup. They were shaping the actual app. "This is real code? Like, if | edit it here, does it - live?" asked Designer 3. That moment, when designers understood that Shaper removes the handoff entirely, was everything. | 1 () il i 1 = Focus on the product Build what customers actually want Don’t get stuck in feedback loops. 0) Oo O (@) 10 7 TM, a Break the traditional way Simple design edits require full dev Dm raracea anAlafeala anne ety Of rena ern2) Ree 1 * SHAPER.STUDIO What's Next By the end of the session, we had pages of notes, videos, and Slack threads full of ideas. And best of all? No surprises. "Every bug that showed up was known. That's the relief. We are on the right track," said Sanket Sahu, our founder. These were not obstacles. They were signals that the structure worked and only needed refinement. Here is what we are already working on: ¢ Fixing snapping, spacing, and height logic ¢ Simplifying the right panel with contextual visibility ¢ Clearer visual indicators for Play vs Edit mode ¢ Better Al feedback — so it’s obvious when it's working - Adding essential shortcuts like multi-select, duplicate, and undo - All of this is gearing up for the next big step: Beta in June. | o Git integration Bridge the gap between engineering and design 0) ] 1 fe) 1 1 0 1 c % O r) Why are handoffs still the norm? 14 Handoff creates friction and ABTA Rca noeco cS leatlesan oO ~@ Shaper Be Part of What We Are Building Shaper is not just another tool. It is a shift in how design and product teams build things together, in real time, with real results. We are building Shaper in the open, with the community guiding every step. If you want to shape what comes next, we would love for you to join us. < Join the waitlist *+ ) Beta launches soon. Be part of the future Scan the QR to join the waitlist Instagram (in) LinkedIn X Trends. Commentary. Experiments. ¥ GeokyAntsX ORF dO tn So lars ey X @geekyants of. Innovate. =< Collaborate. GeekyAnts @ @geekyants Build. BE) (atFollowing Design & Engineering Studio « Pioneering Al Solutions + Impactful Design - Scalable Architecture + Driving open-source Innovation SF - Bangalore » London @ geekyants.com Joined September 2015 98 Following 5,898 Followers Posts Replies Highlights Media GeekyAnts @ @geekyants - 32m es Throwback to April 2012 4+ This was our first-ever work desk featuring the OG lineup — our first MacBook, the first Windows laptop, and, of course, a cameo by the guitar that kept us inspired! ¢ Fun fact: We still have all of these with us today. #TBT #StartupJourney Show mor 17 lOS 26 UXSHIFT ° IOS 26 UX — Needed Change or Marketing Stunt? | WWDC 2025 Breakdown | Amrit Saluja a w Technical Content Writer he core design of iOS remains familiar. However, everything that bloated was escorted out.At first glance, iOS 26 feels ornamental—glass, gradients, motion, tint. It gleams, shifts, and whispers refinement. But polish is not a product. The question lingers: Does any of this make everyday tasks faster? Does it sharpen productivity? Does it make the device more invisible, supporting work instead of interrupting it? Apple calls it Liquid Glass—not because it is smooth, but because it refuses to sit still. UI elements now behave at lightning speed— responsive, refractive, and shaped by movement. Control surfaces adapt in real time. Alerts surface exactly where intent begins. Tab bars melt when not required. It feels organic but surprising. For the first time, the operating system is moving before the user decides. Rakesh Ningthoujam, who leads growth marketing at GeekyAnts, said: “It’s the most user-first update to screens and interaction I've seen. Kind of feels like everyone will copy it a now. And they likely will, not because of aesthetics, but because of the logic beneath the surface. Form Now Follows Anticipation Icons now shift tone with the time of day. Menus respond subtly with gesture. The interface anticipates intent instead of awaiting input. Apple’s design team seems to have traded in static hierarchy for contextual elegance. But beneath the smooth physics, a sharper idea emerges: predictive experience as a UX baseline. This minimalism serves a purpose beyond aesthetics—restraint rooted deeply in system architecture. What does that mean for builders? It means static layouts are obsolete. Products now exist in motion. Interactions must behave predictively. Controls that simply wait for input risk feeling disconnected, while responsive interactions could align more naturally with user expectations. Attention As A Design Principle The Home Screen takes cues directly from content. Images influence how UI elements are highlighted, and Lock Screens adapt typography to complement visual composition. The interface prioritises essentials, eliminating visual clutter. This approach suggests design should not demand user attention but justify it through meaningful interaction. Again, to quote Rakesh: “What Apple has done here is the writing equivalent of editing. They are removing all non-required features and keeping the interactions thoughtful.” The implications extend beyond mere visuals, reshaping how teams might measure success. Engagement metrics might shift from duration of interaction toward efficiency and effectiveness of use. * 10S 26 UX SHIFT Ecosystem-Wide Consistency Apple's recent design choices bring coherence when considered across its ecosystem. Similar principles appear consistently within watchOS, macOS, iPadOS, and VisionOS. What once felt like parallel platforms now feels like a single language, spoken with different dialects. Moving between devices is no longer about replication but about maintaining behavioural consistency. “Universal” now suggests unified interaction patterns rather than just unified codebases. So—Stunt or Shift? Apple has historically avoided frequent visual redesigns; the last significant one was iOS 7 in 2013. Subsequent updates were incremental. iOS 26 represents a noticeable departure. This new design does not primarily seek visual appeal. Instead, it suggests reliability and user- centric predictability. Products operating within this system may now be evaluated by their alignment with user intentions rather than purely aesthetic criteria. Therefore, categorising this update merely as marketing-driven may overlook its deeper implications for user experience standards. ° _ BUILDVSPARTNER ° Does Building an App In- House Actually Slow Launch? ‘a Amrit Saluja Fs a Technical Content Writer Heres what our 800+ development projects tell us — very company starts with a belief. The A Pattern Too Familiar team, the structure, the rhythm— everything seems prepared to take the Most product journeys begin with a bold idea from whiteboard to launch. The decision: to build everything in-house. The team confidence is earned. But too often, the sets out to define every interaction, construct outcome doesn't follow. every foundation, and shape every detail from the ground up. At first, the energy is high. There We have seen it unfold in our 800+ projects: is clarity in the direction and confidence in the bold ambition at the start, followed by talent assembled. Progress is visible, if not avoidable friction at the finish. always fast. As weeks pass, the path grows less certain. Timelines extend. New hires arrive to restore velocity, though the time required to integrate them often exceeds expectations. Coordination becomes heavier. Priorities begin to shift. Despite long hours and sustained effort, the release remains out of reach. This is a common arc in modern product development. It occurs not from a lack of competence, but from a structural overreach that slowly erodes the team’s ability to move == forward. - The shift from working within estat platforms to building entire systerr independently is rarely s | weight. It fragments atte manage compl required to abs soon becon 19 Where Progress Stalls Delivery slows when teams are asked to own challenges that extend beyond their original charter. As responsibilities widen, core systems often remain underdeveloped. Stability becomes difficult to guarantee. Execution continues, but decisions are made in flight, without the grounding of mature frameworks. Scaling, in this state, becomes speculative. Quality suffers as engineers spend more time adjusting their tools than building with them. Each sprint carries more overhead. The product becomes less a vessel of forward motion and more a system of internal upkeep. Tasks remain in motion, but direction fades. The issue lies in the underlying configuration. Many teams are optimised for operation. Far fewer are structured to deliver complex systems at pace. Even the most established companies have encountered this tension. In the early 2000s, IBM transitioned large segments of its non-core IT operations to external partners. This was not a cost-saving measure. It was a strategic choice to regain speed, concentrate innovation, and return internal talent to domains where it could shape outcomes rather than support infrastructure. Well-functioning systems create focus. Poorly framed ones absorb it. The difference rarely lies in effort. It lies in how much of that effort goes toward progress, and how much is spent holding the structure together. * BUILD VS PARTNER Operation Is Not Shipping Most internal teams are structured to ensure continuity. Their work maintains core infrastructure, safeguards data, and sustains system reliability. These are not secondary responsibilities; they are foundational to any organisation's stability. Product development, however, follows a different logic. It operates on shorter cycles and demands a distinct approach—one grounded in speed, experimentation, and clear alignment with user experience. Sustaining a system is not the same as evolving it. To move quickly and build with precision, teams require architectural fluency, tightly integrated feedback loops, and a working knowledge of scalable design systems. This form of execution depends on a different set of instincts. It is not an add-on to operational excellence but a separate mode of thinking—one that many internal teams were never assembled to perform. 20 21 BUILDVSPARTNER ° What the Most Effective Companies Do Differently The teams that succeed at scale are rarely the ones doing everything themselves. They protect internal focus. They let their teams work on the problems that define the product. They bring in expert partners to deliver everything outside the core—faster and with less overhead. One leading U.S.-based insurance company took this path took this path. Internal teams focused on data intelligence and pricing logic, while GeekyAnts delivered customer-facing workflows —modular, API-driven, and fast to ship. This is not about outsourcing to cut costs. It is about staying lean where it matters and accelerating where speed is critical. These companies do not confuse headcount with Capability. They make decisions based on relevance and return. They move forward. Build or Partner? Make It a Strategic Call There is no one-size-fits-all answer. Every product demands a different approach. But strong decisions tend to follow a pattern. Build in-house when the work defines your product, safeguards your intellectual property, or demands long-term flexibility. Partner when speed matters more than control. When you need expertise you do not have. When the task is non-core, but still essential to ship. If your roadmap includes admin dashboards, Stripe integrations, DevOps automation, or one-time migrations, build a product, not an internal agency. External specialists can move faster. And they should. Your team should focus on what only your team can do. BIGO LIVE followed this approach. To move fast in the highly competitive live-streaming market, they partnered with external teams to build and scale their MVP. The result? A timely, successful app launch that captured global attention while avoiding the delays that come from hiring and building in-house. In contrast, competitors that insisted on full internal builds often saw their timelines balloon past relevance. A Final Thought Before You Decide If your launch is slipping, do not ask who is working harder. Ask who should be working at all. In-house is the wrong default. Look at the projects that matter. Map what gives you an edge. Protect your team's time for that, and make room for trusted partners to take on the rest. We help companies find that balance. If you are trying to decide whether to build or bring in help, we can give you a roadmap built from experience. for your expertise. Geek Talksand GeekyAnts Podcast S totesf till Obes a The GeekyAnts Podcast oO = innovation \nsicers The GeekyAnts Podcast GeekyAnts GeskyAnts w No rating « Technology i ee Episodes About Morelike this Allepisodes - Newest Does Al Really Make Better Decisions core 1hanHumans?|GeekyAnts Podcast In this episode, we explore one of the hottest questions in tech: Can Al truly make better decisions than human... 16 Oct 2024 « 6min © © < : © We =Whatis TLV Objects? Where And were Why Do We Use Them| Tech Discus... fl @, ID Search Your Librarv FEATURE ARTICLE ° No More Spreadsheets: Meet the Al Estimation Tool for Enterprise Projects nterprise projects often miss the mark - before a single line of code is written. The fault lies in estimation, still largely driven by spreadsheets, inconsistent inputs, and decisions shaped by guesswork. Nearly 66% of software projects overshoot their budgets. With global software spend projected to reach $1.25 trillion by 2025, this margin of error carries unacceptable cost. At GeekyAnts, the goal was to address this problem at its root. The result is the Al- powered Estimation Builder—a tool that delivers accuracy, speed, and structure to the earliest phase of software development. Built with generative Al and trained on years of internal project data, it is designed to help enterprises scope work with greater clarity and less friction. From Hackathon Spark to Internal Platform The idea for the Estimation Builder came from within GeekyAnts. “We needed a faster, smarter way to scope projects—one that did not depend on five follow-ups,” recalls Sanket Sahu, Founder & CTO (Innovation) at GeekyAnts. The estimation process had become bloated: spreadsheets riddled with fragile formulas, email chains between sales and tech teams, and constant cross-checking. It worked, but only just. That friction led to a hackathon prototype using OpenAl's API. Engineers at GeekLabs built an internal system that combined client inputs and past project data to generate effort estimates and recommend feature sets. What started as a technical experiment evolved into a full-scale tool. GeekyAnts Launches AI-Powered Estimation Builder to Help Enterprises Plan Software Projects Smarter June 10, 2025 2:15PM / “_ Digital Team Estimation In Minutes The Estimation Builder uses predefined templates for common app types like fintech and e-commerce to help teams get started quickly. It draws from internal knowledge and ChatGPT to recommend features based on the project type, using agile estimation logic to group features by complexity and readiness. Assumptions, specific requirements, and external links can all be added ina single interface. Once the scope is defined, the system instantly calculates a realistic cost and timeline based on similar past projects. Output can be exported in CSV format and supports multiple currencies (USD, INR, GBP). Key Capabilities And Impact The Estimation Builder is built for speed, precision, and collaboration: ¢ Centralised Workspace: Estimations are stored in a single portal—version control is no longer an issue. ¢ Al-Supported Planning: Integrated intelligence recommends features based on industry patterns and prior delivery. The system learns and improves with each use. ¢ Fast Turnarounds: Estimations that once required days of coordination can now be completed in hours. ¢ Higher Accuracy: Automated logic replaces manual calculations. Estimates reflect real project baselines. ¢ Reusable Templates: Pre-built modules help fast-track proposals across verticals, reducing reliance on senior engineers. ¢ Collaborative Workflow: Sales and tech teams can align and adjust estimates together in real time. Exportable Output: Estimates are ready for presentation or integration, and can adapt to global pricing models. * FEATURE ARTICLE Results Since Rollout In the 18 months since its internal deployment, GeekyAnts has used the Estimation Builder to generate more than 150 structured project estimates. The pre-sales team has developed over a dozen industry-specific templates, reducing manual coordination and boosting client response times. Engineers have added over 30 enhancements to the tool based on ongoing feedback. The workflow impact has been immediate: turnaround time for proposals has dropped sharply, and estimation quality has become more consistent across projects. A Broader Al Roadmap The Estimation Builder is part of GeekyAnts’ growing suite of Al-driven tools: ¢ Shaper: A collaborative platform where design and code coexist. ¢ gluestack: A streamlined development framework for web and mobile. ¢ Hiroscope and AntEngage: Al-based recruitment and IVR tools already in production. What Comes Next New features in development include a document-parsing assistant that will extract scope directly from briefs and generate structured estimates. Integrations with platforms like HubSpot and Jira are planned to automate estimation workflows. A visual analytics dashboard is also on the roadmap. Long term, GeekyAnts is evaluating a SaaS release of the Estimation Builder, opening access to other service providers and enterprise teams. The tool is now available to key clients and partners. Product leaders and CTOs interested in transforming their estimation process can reach out to GeekyAnts for a walkthrough. 24 25 REACT NATIVE PROGRAM TOWNHALL ° Lessons from the Stack: What Townhall Taught Us About Real-World React Native Sarthak S BakreJoydeep Nath 2 Rajat Chaudhary 4 Software Engineer IITech Lead I] if Software Engineer Il his edition of the React Native Townhall centred on practical insight. Developers shared grounded stories drawn from production work—scenarios shaped by constraints, ambiguity, and the need for repeatable solutions. The session prioritised implementation over abstraction, with a clear focus on what it takes to build well in real environments. MEAL INSPIRATIONS Stronger Declarative Navigation: Simplified, Not Weakened Anubhav presented updates from a food ordering app where the navigation stack had grown unpredictable. Features like table booking, offers, and menus each launched into separate flows, and users could jump between them from several entry points. The original implementation used deeply nested React-Navigation stacks. Debugging became difficult as the logic sprawled across components. The refactor moved logic to a single navigator.ts file and introduced a context layer to track routing state. What made this work was not just the rewrite. It was the use of a clear mental model. Navigators were no longer coupled to screen logic. They existed as first-class modules with exported functions. These functions, like navigateToOffers()or openMenuScreen(), became stable entry points for the rest of the app. This allowed testing, refactoring, and even mocking navigation flows without re-rendering UL. It also made it easier to onboard new developers, who could understand app routing by reading one file rather than tracing nested trees. “mmeg ~S@- * REACT NATIVE PROGRAM TOWNHALL Handling Offline Constraints Gracefully Srikant walked through the integration of React- Query for a warehouse management app. The system required scanning inventory, updating stock, and syncing offline actions when the device reconnected. The naive approach had been to build custom polling and reconciliation logic. This quickly became unmanageable. The team then moved to react-query with cache persistence enabled. Mutation queues were created using React Query’s built-in mechanisms. When network connectivity was lost, all new mutations were written to disk. Upon reconnection, react- query rehydrated the cache and dispatched the queued mutations in order. No external queueing library was used. The cache became the single source of truth, both for UI rendering and retry logic. The implementation also avoided introducing a Redux store, reducing overhead. It was an elegant fix that respected the constraints. import { QueryClient, QueryClientProvider } from ‘@tanst import { ReactQueryDevtools } from ‘@tanstack/react-quer const queryClient = new QueryCLlient(); root.render ( client={queryClient} React Query REACT NATIVE PROGRAM TOWNHALL ° Minimal Permissions, Maximum Utility Another highlight was shared from a health- tracking app where the client had specific privacy restrictions. No unnecessary permissions could be requested, and all user interaction data had to remain device-side. The problem: tracking screen transitions and button taps for workflow analysis. The solution: a custom InteractionLogger class that captured events and wrote them to encrypted storage. No third-party analytics SDKs were used. Events like screen_view, button_click, and form_submit were stored as JSON with timestamps. A scheduled background task, using expo- task-manager, uploaded logs only when the user explicitly enabled diagnostics. Until then, all data remained offline. This allowed the team to implement behavioural insights without breaking the privacy promise. iPhone 12 Pro Max iOS 14.5 Location updated! 4 {"appState":"background","headless" :true,""secureStore Val":null} Shortcuts Contacts LegendApp/ legend-list A high-performance list component for React Native RB 20 Legend Kit Under the Microscope Performance surfaced repeatedly. Sarika shared findings from evaluating LegendList as a potential upgrade over FlashList. In apps with dynamic headers, multiple scroll axes, and sticky categories, FlatList had been reaching its limits. LegendList allowed the team to decouple the scrollable region from the rendering container. This enabled smoother transitions and allowed animations to be run outside the main thread. But with flexibility came fragility. Scroll events occasionally desynced when two nested lists were used. The workaround was to enforce a scrollSyncController that explicitly linked parent and child lists. This kept gesture priorities aligned and resolved jitter. The improvement was measurable—scroll FPS improved from 48 to 58 on mid-range devices. Still, the team concluded that LegendList works best when planned from the start. Retrofitting it into apps designed around FlashList introduced hidden coupling that required cleanup. G2 306 W 8 w 2k ¥ 64 Used by Discussions Stars Forks Reusable UI, without Overhead Deepak presented a component library built for three client projects in the education vertical. The challenge was to keep the library lean. Shared components had to avoid assumptions about fonts, spacing, and data shapes. The team used a local monorepo setup. Each component was written in isolation using Storybook, then consumed via path aliases. The key: no component in the library could import client code. All props had to be passed explicitly, and defaults were avoided unless required by design tokens. This led to a stricter, more verbose API—but also made the components more portable. Over 70 percent were reused across apps without modification. @@@ = & confident-breeze-7mnsr-Coc X + <€ ie File Edit Selection View Go Help ry) refine Project i= Posts > CG Q 4 codesandbox.io/s/confident-breeze-7mnsr?file=/src/App.js @®aQrk @ks A (=) H ie) & Share Y Fork ~ CreateSandbox & : Js Appjs e moms Browser st import { useForm } from "react-hook-form"; C https/7m il 2 3 let count = 0; is) React Hook Form 5 export default function App() { const onSubmit = (data) => alert (JSON. stringify(data) § const { register, handleSubmit } = useForm(); First Name return (

Console @ Problems @ Ln11,Col21 Spaces:2 UTF-8 LF JavaScript 0 Employed * REACT NATIVE PROGRAM TOWNHALL Voice Input, but Selective From a logistics app, another team demonstrated how they integrated react-native-voice for multilingual voice input. Drivers could dictate package notes in their local language. But react- native-voice only supported one language per session. To solve this, a language picker was introduced before the microphone was activated. The app then calls Voice. start(selectedLanguageCode) to initialise the recogniser in the chosen language. Recognition accuracy improved significantly. The default English fallback had been discarding up to 30 percent of notes. With explicit language selection, that rate dropped to under 5 percent. The Real Work Is in the Details Many contributions focused on engineering hygiene. One team cleaned up deep prop-drilling by adopting context providers, limited to small areas of the app. Another shared how they used react-hook-form with schema-based validation to reduce form errors. Across the board, the improvements were small, but intentional. The townhall ended without ceremony. No wrap- up, no call to action, only a slide reminding contributors to add their updates to the shared deck before next month. That simplicity is the point. The React Native Townhall has settled into its rhythm, driven by code, care, and the discipline of consistent practice. The Podcast Idea 45:00 dd Add music LEGALUPDATE ° 4. Monday. Jun 8 cue 76 Issue if # cityaailynew® From Policies to Petitions: June’s Legal Snapshot Christine Mathias Manager Legal he legal and compliance operations at GeekyAnts remain focused on ensuring continuity, transparency, and readiness for cross-border engagements. While certain compliance activities are still underway and will be detailed later this month, the legal team has shared a summary of key updates from June. Insurance Renewals in Progress The renewal process for both Professional Indemnity and Cyber Security Insurance is actively underway. These coverages form a core part of the organisation’s risk management framework and are being reviewed to maintain adequate protection across service lines and geographies. Visa and Mobility Updates H-1B visa filings have been initiated for five candidates who are being onboarded under GeekyAnts Inc. The documentation and submission process is being handled under United States Citizenship and Immigration Services (USCIS) timelines and requirements. Separately, the L-1A visa petition for Saurabh has reached its final stages. This petition supports his upcoming assignment in the United States and is expected to be filed shortly, following final legal checks. ° LEGAL UPDATE Certification and Policy Updates Work is in progress on the ISO certification renewal, with relevant processes, documentation, and audit checkpoints being finalised. This effort reinforces the organisation's commitment to security, quality, and compliance. A new Privacy Policy has been published on the GeekyAnts website. The updated document brings greater clarity to how user data is collected, stored, and processed. In parallel, revised Terms of Use are being prepared. Once finalised, they will replace the current version and apply to all users accessing services via the official website. Contract and Document Review In June, the legal team conducted a detailed review of more than 20 client agreements and 25 supporting documents. These reviews ensured alignment with business intent, legal integrity, and jurisdictional requirements. The process involved negotiation support, clause interpretation, and final approvals for execution. DESIGN SPRINT RECAP ° Inside a High-Stakes Design Sprint Chetan Hegde UI/UX Design Lead II orking on a fast-paced project with Applying Structure to Meet Pace fixed timelines, the objective was clear from the outset: deliver a The team comprised five designers working in solution that aligned with both user needs and close coordination with cross-functional the client's business priorities. Design and partners. The first week was dedicated to development moved in parallel, which made aligning with stakeholders—understanding their early clarity and consistent coordination vision, values, expectations, and working styles. essential. To maintain momentum, the team set up daily syncs to stay coordinated and surface blockers early. JIRA cards were maintained from the start to bring clarity and accountability across all moving parts. Design and engineering collaborated from the beginning, validating feasibility before any designs were finalised. Mind mapping helped unpack complex or ambiguous flows before progressing to wireframes. A modular design system was introduced to ensure both visual and functional consistency. Weekly design sprints structured the work, while clear documentation supported smooth downstream development. Throughout the project, the team operated in short feedback loops, prioritising quick iterations and alignment over polishing in isolation. Delivering with Confidence The work was delivered on time, without compromising on user value or business needs. Every decision was shaped by shared understanding, structured execution, and early collaboration, enabling the client to move forward with clarity and confidence. We Share. We Geek Out. Catch the brains behind the builds, stories from the trenches, and everything in between—on Linkedin. qe Q GeekyAnts GeekyAnts 9 Design & engineering studio with a user-first and Al approach. IT Services and IT Consulting * Bangalore, Karnataka 136K followers * 501-1K employees Home About Posts Jobs’ People €& Images GeekyAnts 135,658 followers 23h: ® Discover how Al is transforming industries and Videos Articles Innovate. Collaborate. Build. a = Trim) © Insigh' Dc + Follow Linked [i 33 AGENTIC Al IN FINTECH Sanna Samvedana Bara Community Associate orget speed—autonomy is the new benchmark in fintech operations, and Agentic Al is what's powering it. As digital workflows scale, systems must do more than automate—they need to act, adapt, and sustain themselves.. Agentic Al answers this need by reengineering the decision-making layer within financial operations. Tasks are executed. It brings autonomy into decision-making systems by embedding intelligence directly within operational flows. These systems function as discrete units within the process —executing tasks, adapting to inputs, and preserving continuity without routine supervision. In financial environments where scale, compliance, and latency are critical, this model enables process orchestration with speed and accuracy. Tasks like risk detection, transaction tracking, approvals, and reporting now run faster, with more consistency and less manual drag. Boudhayan Ghosh Technical Content Writer Operational Impact In large-scale financial ecosystems, the addition of agentic layers is producing measurable shifts. Mastercard implemented generative models across fraud detection pipelines. This led to a substantial increase in detection rates and a marked reduction in false positives across billions of transactions. Banks and financial institutions are investing in similar systems to stabilise key functions. JP Morgan Chase has applied agent-based automation to streamline compliance workflows and internal audits. Affiniti and Hyperbots are developing product lines grounded in agentic operations, supported by early-stage venture funding. These systems are being adopted not as supplements, but as critical infrastructure within existing environments. Workflow Architecture The application of agentic Al within risk and fraud pipelines simplifies multi-step processes. A typical loan lifecycle—spanning identity verification, risk scoring, and compliance documentation—can now be executed by coordinated agents across modular checkpoints. In internal trials, the end-to-end cycle duration was reduced from several hours to under one. Fraud investigation workflows, once dependent on analyst intervention, are now managed by autonomous agents. These systems monitor real-time transactions, generate flags, and produce audit logs automatically. Tasks that previously required full-day cycles are being resolved in near real time, with fewer escalations and lower resource expenditure. Deployment Guidelines The introduction of autonomous agents requires deliberate structure. Financial systems must maintain precision across all functions, from user onboarding to regulatory compliance. Successful deployment depends on several factors: ¢ Defined workflow logic Systems must align with existing process sequences. Open-ended tasks do not yield stable outcomes. * AGENTICAIIN FINTECH ¢ Reliable data architecture Clean and accessible datasets are foundational. Agents must operate on valid, version-controlled inputs. ¢ Monitored execution environments Logs, audit trails, and observability tools are necessary for both internal confidence and regulatory review. ¢ Scoped responsibilities Each agent must operate within a well-defined boundary. Expanded capability comes only after baseline performance is validated. Training and adaptation are also part of the rollout process. Operational teams must interact with these systems not as users, but as collaborators —overseeing results, refining triggers, and ensuring alignment with business rules. A System Designed for Continuity Agentic Al introduces a structured approach to automation in finance, aligning intelligent systems with operational continuity. It unifies processes that often rely on disconnected tools and conditional pathways. Through embedded intelligence, it enables consistency across workflows, reduces manual intervention, and supports scale without compromising control. This architecture is being implemented across core financial functions—loan evaluation, support systems, compliance, ledger operations, and risk modelling. Each integration adds clarity to execution and strengthens the system's ability to respond to high-volume, high-frequency tasks with precision. As these implementations mature, agentic frameworks are beginning to define the architectural standard for building and maintaining financial technology. Antagenis Agentic Al Powered Business Process Orchestration Automate and manage enterprise-level processes with a colony of Al Agents that think, act, and evolve. Schedule a Demo Start Turning | _ Silos to Synergy : Get your.army of AntAgents today. __ * USAIMEETUP New Architectures of Work: Voices from the US Al Meetup Understanding the Shift from Prompts to Performance Takasi Venkatesh Sandeep, Tech Lead at GeekyAnts, leads a team focused on developing Al-native tools and decision-making systems for enterprise use. Over the past year, he has concentrated on building intelligent agents that do more than respond to prompts. These systems are capable of executing tasks, making decisions, and collaborating across functions. While chat interfaces and retrieval systems have dominated the first wave of generative Al, Sandeep's focus has shifted to structured agents operating within complex environments. These are systems designed to carry out workflows, not Just conversations. An Al agent is fundamentally different from a chatbot. Rather than offering a single response to a question, it engages in multi-step tasks, drawing from memory, accessing external tools, and adapting its behaviour based on feedback and outcomes. Each agent is anchored by four components: a defined persona, a memory layer, an integrated toolset, and a reasoning model. Sandeep illustrated how these agents are already being applied in production. In software engineering, for example, agents can monitor error reports, trigger alerts, and even draft support tickets. In healthcare, they are used to handle appointment bookings and clinical updates. These implementations are not speculative. They require precise orchestration of models, APIs, and business logic. At GeekyAnts, Sandeep's team uses LangChain, along with LangGraph and LangSmith, to build and manage these agents. In one prototype, three agents worked in tandem: one to parse requests, another to assess technical feasibility, and a third to prepare structured responses. This architecture mirrors how teams collaborate, but with the advantage of immediate responsiveness. 3/7 USAIMEETUP ° Successful implementation starts with defining the process, not the model. The team identifies workflows, maps tools, labels decision points, and builds agents with clear, scoped responsibilities. Memory systems retain relevant history, enabling agents to evolve based on previous interactions. The enterprise shift toward agentic systems is already underway. In sectors like healthcare, finance, and automotive, companies are deploying agents for diagnostics, compliance, and inventory planning. These systems are becoming part of long-term infrastructure, not experimental pilots. However, they require structure. Poorly scoped goals and vague logic can lead to failure. Sandeep emphasises the importance of human oversight, logging, and feedback loops to ensure agents remain reliable and aligned with business goals. Teams are advised to begin with one repeatable workflow, test for stability, and scale only after results are validated. Tools like LangChain, AutoGen, and AWS Bedrock offer the necessary building blocks for these systems. With intent and clarity of purpose, teams can begin small and grow with confidence. Design, Code, and the Builder Shift Sanket Sahu, CTO (Innovation) at GeekyAnts, has spent his career building tools that bring design and code closer together. His work includes open-source platforms like NativeBase, BuilderX, and gluestack. Today, his focus is on restoring creativity to software development through Al and eliminating the friction between designers and developers. Sahu’s interest in visual programming began in the late 1990s with tools like Visual Basic and Macromedia Flash. These environments allowed developers to create user interfaces directly, with logic layered underneath. That ease faded as mobile platforms evolved, and the divide between design and code widened. By 2025, tools like Figma and VSCode still operate in separate silos. But Al is beginning to close the gap. Through what developers call “vibe coding,” natural language prompts now generate working prototypes, games, and entire applications. Real-time development, once the domain of simple tools, is becoming viable for production use. Two shifts define the current landscape. First is augmentation: tools like Cursor and Firefly integrate Al into existing workflows, increasing pace without changing habits. Second is replacement: platforms like Locofy and Builder reduce the need for traditional development pipelines, enabling small teams to scale efficiently. At GeekyAnts, this shift is supported by tools like AntAgents and Shaper. AntAgents map Al personas to business roles such as sales, QA, and design, with humans validating outputs. Shaper, on the other hand, merges development and design environments. It opens a Next.js project as a design file, allowing visual edits to reflect instantly in code. This merging of roles points to a broader transformation. The boundaries between designers, developers, PMs, and founders are softening. A new kind of role is emerging— one Sahu refers to as "the builder"—marked by fluency across disciplines and a closer relationship between thinking and execution. Bhushan S 5 rahulkumar N Ashutosh kunalshivam aashikothari Vikas Kumar Rohit G [ Nischit Shetty) Aman Fangeria Prince thakur Jaspreet Singh FINTECH CASE STUDY ° Building Scalable Fintech: A Global Payment Processing Success Story ‘1 he Boudhayan Ghosh Technical Content Writer ur client is a North American payment processing company that wants to create a world where every individual and business alike can effortlessly manage their money. Their goal is to create payment solutions that are not only secure and accessible but also empower people to achieve greater financial freedom and connection. Overview We developed a solution that enables seamless management of global transactions and facilitates expansion into new markets through an intuitive admin panel. This system allows for the addition of new destination countries and the provision of services without reliance on technical teams. The app has boosted the user base across multiple regions. It can process a significant volume of daily transactions, demonstrating its efficiency and reliability in delivering financial services on a global scale. 120K+ Active Users in the UK, Canada, Europe, and Australia 400M+ Global Payments Processed Annually 350K+ Downloads across iOS and Android ld all wa ile Business Requirement The client aimed to address challenges and provide customers with more efficient options for sending money to their payees or beneficiaries. We needed to achieve their vision to transform how people manage their finances by prioritising seamless and secure solutions. Their focus was on delivering intuitive payment services that emphasised both security and efficiency, ensuring an improved customer experience. * FINTECH CASE STUDY Solution We suggested a comprehensive solution that allows users to track all their transactions, receive global notifications from the admin, view news, and easily send and receive money with other users in the same region. With built- in currency exchange calculations, users can enter an amount and instantly see the equivalent in the destination country’s currency. Core features such as account management, profile settings, notifications, updating transaction pins, and support were included to provide a complete user experience. For the admins, we proposed full control through an intuitive admin panel, with the following key features: - Features can be toggled on or off as needed. - A lightweight CMS to allow easy updates, such as changing banners within the app. ¢ An informative dashboard for monitoring transactions. - Detailed user activity and listing insights for better platform management. FINTECH CASE STUDY ° Challenges In response to our client's fast-paced project ¢ Minimised interservice API calls and needs, we implemented several technical eliminated redundant code in backend solutions to ensure seamless development, microservices by implementing a monorepo scalability, and team collaboration. By approach. addressing challenges such as mobile app deployment, backend optimisation, and *» Maximised CPU efficiency and handled testing, we were able to enhance project heavy user loads by enabling ECS auto- efficiency and maintain high-quality standards. scaling, implementing Redis, and reducing network calls with caching. « Enabled the creation and publishing of two mobile apps from one code base using App *- Reduced QA time for testing builds in four Centre pipelines and AWS secrets for regions by automating end-to-end user dynamic environment switching. flow testing. | - © 0% 0 @@ gore Our Approach We fully analysed the requirements and roadmap to provide the client with a scalable solution that reduced reliance on time-sensitive tasks by admins. This approach enabled expansion into new countries and allowed us to quickly remove features that underperformed or faced compliance and third-party contract challenges. We achieved this through Agile methodologies and strong project management with an optimal team size. Backend Development We have implemented a microservice architecture using six Node.js services and are currently migrating from a multi-repo setup to a monorepo using npm workspaces. This migration is designed to enhance code reusability and reduce inter-service API calls. Our infrastructure is hosted on AWS ECS, and we utilise various AWS services to ensure optimal performance, security, and efficiency. Key tools and configurations include: + Storage: Images and documents are stored in an S3 bucket, with IAM user access configurations for security. * Caching: Redis is used for caching database query responses, improving API response times. ¢ CI/CD: Bitbucket Pipelines manages backend service deployments. + Testing: Unit testing with Jest, load testing with Grafana K6, security testing using OWASP tools. FINTECH Frontend Development Our front-end architecture leverages caching and automation tools to optimise performance across both web and mobile applications. By integrating robust hosting and deployment pipelines, we ensure efficient and seamless app management for our users. Key tools and configurations include: « Caching: Redux Persist is used for front- end caching in both web and mobile applications. ¢ Hosting: Web apps are hosted on AWS ECS. « CI/CD: Mobile app pipelines are managed via App Centre, and web app deployments are handled using Bitbucket pipelines. - Testing: Unit testing with Jest, automation using Appium and Playwright 42 FINTECH CASE STUDY ° Project Execution We ensured that all requirements were thoroughly analysed, clarified, and finalised before beginning design or development. A detailed Technical Requirement Document (TRD) was prepared to capture all necessary technical details, and we planned the execution timeline across sprints, keeping security a top priority. Daily scrum meetings and sprint planning helped monitor team progress and capacity, while requirement gathering calls with the client ensured full feature capture. Transparency with the client was maintained through regular updates, thorough testing, demos for feedback, and post-release monitoring, with a focus on resolving any revenue-impacting or compliance issues. Final Deployment Over the past six months, we fixed the CPU utilisation overflow issue, maintaining usage at a consistent 58% post-refactoring. The performance of backend services, the admin panel, and mobile applications (PayPenny & MoneyFirst) has been enhanced. We've also improved the maintainability and reliability of the codebase while reforming the infrastructure to boost overall application performance. Security has been strengthened with WAF- firewall protection against DDOS, injection, and CSV attacks, along with IP whitelisting and restricting access by country for both the admin panel and mobile apps. Key features were reformed to reduce reliance on the tech team, allowing the admin more control. Tools like Centric, Mixpanel, Uptime Robot, and AWS | monitoring systems were integrated to track and resolve production bugs efficiently. | Project Results Since December 2020, the app has recorded 100K Android and 250K iOS downloads, with active user numbers of 44,396 in Canada, 39,814 in the UK, 26,851 in Europe, and 8,961 in Australia, indicating significant adoption in these regions. The app processes around 500 deposits and 300 withdrawal transactions over 3 days. Specifically, Canada averages 512 deposits and 383 withdrawals, the UK 611 deposits and 335 withdrawals, Europe 555 deposits and 342 withdrawals, and Australia 107 deposits and 168 withdrawals, demonstrating steady transaction activity across all regions. 120K+ Active Users in the UK, Canada, Europe, and Australia 400M+ Global Payments Processed Annually 350K+ Downloads across iOS and Android Life at The Geek Base < geekyantsofficial GeekyAnts 721 2,163 39 posts followers following ~—— Product/service N Product studio with a user-first and Al-driven approach.Innovate. Collaborate. Build. GEEKY ANTSOFFICIAL Posts Cay geekyants.com and 2 more e geekyantsofficial 49% Followed by sailesh_ —, Vr and 14 others | I i] } ¥ Yeekyantsotficial and Qluestackio IEA —Sann_day ¥-GeokyAnts Dev Stories - Reels to YO... Dev Stories Articles | Tf oup office i= 9Poup chat gets & Py F. US arresteg ary EXW Liked by lone_meanderer and oth kyantsofficialThe first React and 5es... < SeeKy of 2025 was nothing shortof a oe Re games, meaningful collaborations, engagin 20 January + tal ntenfficls XP noolnis © a & B 8 @geekyantsofficial ~ AIPROGRAM TOWNHALL * Townhall Boudhayan Ghosh Technical Content Writer he recent townhall presented a series of grounded demonstrations, each tracing the operational layers of Al systems in production. The focus remained on integration—how intelligence is structured, how workflows are shaped, and how infrastructure choices influence delivery. Across three segments, the work revealed a consistent approach: building Al not as an overlay, but as a function embedded into system logic, shaped by constraints, and optimised for real-world behaviour. 45 Inside the Al Program The first demonstration featured a collaboration with BHP PowerTech, where the goal is to streamline access to enterprise data through a conversational interface. The problem is familiar: a business team needs answers, a data team holds the key, and the translation between them is slow. The solution is newer. A chatbot, layered atop a Snowflake warehouse and powered by a large language model, is being trained to respond to business questions in natural language, turning SQL queries into insight and interface. ata security policies. es in orchestration. aWws ~~) Summarize customer sentiment * AlPROGRAM TOWNHALL The second demonstration showcased a nutrition tracking system designed to process inputs across multiple modalities—voice commands, text entries, and image uploads. What began as a tool for single- meal logging has evolved: users can now add multiple entries at once, with the interface prompting intelligent choices about whether to merge or replace. Under the surface, optimisations have been made with care. APIs have been tuned to fetch data in fewer calls. Payloads are leaner. Prompt struc have been reduced to avoid latency withou compromising clarity. These optimisati but they are only part of the story. T significant advancement lies in coor includes how the system decides tool to invoke, howit filters irrele how it adapts responses basec Intelligence in this case dep the model knows, but onF oy +7 VJ AlAgents + D Sales Agent + [DD Help Desk Agent LD Query Agent salesforce 46 AIPROGRAM TOWNHALL °¢ The third segment of the townhall focused on infrastructure innovation through a proof of concept built on Snowflake Cortex. Unlike traditional Al architectures that rely on fragmented services for data storage, model execution, and vector search, Cortex consolidates these capabilities within a unified environment. In this POC, data remained entirely within Snowflake, eliminating the need for external transfer. Prebuilt models were executed natively, and vector operations, such as embedding- based search and retrieval, were performed without relying on third-party tools. Complex workflows, including document parsing, chunking, and response generation, were implemented using standard SQL and lightweight Python. The demonstration prioritised control and clarity. A full conversational agent was constructed, with search pipelines and LLM prompts executed entirely inside Snowflake. Rather than focusing on scale, the emphasis was on simplicity, speed of integration, and reduced operational overhead, demonstrating how infrastructure can enable performant Al systems with minimal friction. As the session drew to a close, the conversation widened. New frameworks entered the mix: Google's agent-to-agent protocols, OpenAl’s Codex suite, and open-source benchmarks that now edge past Claude 3.5 in core engineering tasks. As the session drew to a close, the conversation widened. New frameworks entered the mix—Google's agent-to-agent protocols, OpenAl’s Codex suite, and open-source benchmarks now edging past Claude 3.5 in core engineering tasks. The session moved with focus. Each demonstration circled the same set of concerns: integration, orchestration, and the realities of deployment. Across tools and interfaces, the work stayed close to how systems behave under pressure—how they handle input, resolve uncertainty, and align with existing workflows. The positive? The direction is taking shape through the work itself. Take the Edge With You Read Al Edge Magazine on the go Tap into the future. One scroll at a time. @GeekyAnts ACCOUNT MANAGEMENT ° From Pipeline to Praise: Highlights from the AM Frontlines AnnalisaJ Urumbath _ a Jyotsna Senior Account Manager Account Manager he AM team at GeekyAnts is driving forward with renewed energy, marked by new client wins, strong delivery . outcomes, and early praise from partners. The past few weeks have been packed with critical progress across design, development, and deployment phases, each backed by encouraging signs of client satisfaction and trust Building for Scale Interviews are underway to onboard Java developers for an upcoming banking sector initiative focused on creating a standalone, cross-platform app for both mobile and web. The engagement is still in early stages, but expectations are high. If secured, the project is expected to deliver strong business value and establish a long-term collaboration. A Promising Start The sales team has successfully converted a computer vision startup into a new client. The engagement is currently in the design phase, with clear potential for scale depending on the team's early performance. It is a promising partnership with room to grow. Delivered Successfully Two major projects—a digital health platform and an education-focused retailer—have been delivered successfully within their stipulated timeframes. The teams involved demonstrated discipline, alignment, and execution consistency across the board. The digital health team was rewarded with a 5- star Clutch review, further validating the project's quality and impact. * ACCOUNT MANAGEMENT Commended for Craft and Care Positive Signals Ahead of Launch The client behind a smart security app shared With deployment planned in the coming heartfelt appreciation for the project team’s month, the client behind a clean-tech mobile work, noting their “attention to detail, app has already expressed their enthusiasm. In professionalism, and the way they've particular, they are “really impressed with the approached each stage of the project.” They design and UI"—a strong vote of confidence added, “We're feeling confident about the ahead of launch. progress and excited for what's coming next.” A strong endorsement for the collaboration and care shown throughout. Devices Sensors/Devices in your system Shaping Al Conversations In arecent client call that included the GeekyAnts CTO, a developer tooling company expressed keen interest in integrating Al into daily developer workflows. They also asked the team to guide them on the responsible usability of Al, underlining how Al is increasingly becoming a strategic theme across engagements. French Doors co Door Lock Sensor CGB Unlocked & Closed Was Front Door Door/Window Contact Kitchen Window Window Sensor Side Door Door/Window Contact Elements StoreAbout i: #35363A; dimage { none; . 1008; /design ‘--events: none; on: fixed; ; ility: hidden; +56 ey a sz 100%; +7~background-image ] #backgroundImage { sibility: visible; os OM ss 3: o1 DevOps, Rewritten: From Tools to Systems Thinking Reimagining Cloud Architecture with GenAl Amoghavarsh Patil, a cloud architect and engineer, built Cloud Sketcher to address a recurring frustration: the time and effort required to design, translate, and document cloud architectures across AWS, Azure, and GCP. With consulting demands requiring rapid cross-cloud comparisons, he began exploring ways to streamline architectural workflows using generative Al. Cloud Sketcher emerged as a tool built from real-world needs. It allows architects to sketch once and generate platform-specific diagrams, documentation, and even Terraform code. Its modules handle visual conversions, generate cost estimates, and soon will offer deployable infrastructure as code. Unlike generic diagramming apps, Cloud Sketcher is tailored for cloud engineers. It comes preloaded with over a thousand cloud-native components and emphasises accuracy over aesthetics. Adoption has been swift, with over 2,000 users in just three months. Engineers from companies like IBM, Bosch, and Aspire have incorporated it into their design routines. The platform reflects Amoghavarsh’s belief that the tools we use should accelerate architecture without compromising clarity or consistency. From DevOps to Platform Thinking Turja Narayan Chaudhuri has spent over a decade shaping infrastructure systems and developer workflows across industries. Today, he leads a global platform engineering team and participates actively in shaping platform standards through CNCF and the DevOps Institute. His reflection traces the journey of DevOps from a cultural movement to a tool-centric implementation. While adoption has grown, disparities remain—only a minority of teams consistently meet benchmarks like deployment frequency and recovery time. Turja sees internal developer platforms as the next step. These platforms abstract shared tools and services into reusable building blocks that internal teams can access securely and consistently. In practice, his team began building such a platform in 2018, now serving developers across geographies. Their goal is to reduce cognitive load, not just technical debt, by centralising observability, security, and deployment tooling. Turja positions platform engineering as a maturity model, not a replacement for DevOps, and encourages experienced engineers to build internal systems that empower others to deliver software more efficiently. Inside the Black Box: Observability for Microservices Mithun K has spent the last two years focused on observability within microservices at GeekyAnts. His work highlights the growing complexity in modern distributed systems and the role observability plays in reducing ambiguity during incidents and performance slowdowns. He distinguishes between monitoring and observability, framing the latter as a way to trace root causes rather than surface-level symptoms. His setup leverages OpenTelemetry with Prometheus, Loki, and Tempo to collect metrics, logs, and traces in a unified system. The advantage lies in correlation—being able to follow a request across services, identify a problematic query, and resolve the issue with clarity. In one project, Mithun’s team resolved a bottleneck caused by a slow SQL query only after they introduced tracing. The experience underscored the need to instrument early, before performance issues arise. OpentTelemetry offered them backend flexibility and consistent data flow across services. For Mithun, observability is not just about tooling—it is a way to build engineering confidence through visibility. Bharath Nallapeta manages infrastructure with » =, At Marant’s Open Source Program Office a " a focus on upstream contribution and poe sustainable operations. After years of working with Kubernetes, he began to see recurring challenges in scale, consistency, and observability. The solution, for him, was not another tool, but an internal platform. His team built CORD—an open-source Kubernetes platform designed to simplify provisioning, automate observability, and centralise state management. It is powered by modular components like KOs, Cluster API, VictoriaMetrics, and OpenCost. CORD enables developers to request environments quickly and maintain consistent observability across all clusters. One of its standout features is a shared service catalogue that teams can access with a single Helm command. Bharath's approach reflects a belief in design over accumulation—building infrastructure that scales sensibly, not endlessly. His team encourages others to test CORD themselves. It runs locally, supports Raspberry Pi clusters, and is transparent by design. The intent is not to hide Kubernetes but to make it easier to live with. DEVOPS MEETUP ° A Mindset for Cost-Efficient Cloud Engineering Aditya Prakash views cloud efficiency as a cultural priority. In his consulting experience, cloud bills are not just technical feedback— they influence client relationships, renewals, and business continuity. His approach avoids shortcuts. Instead, he emphasises architecture choices, code efficiency, and well-scoped observability. One case he recalls involved poorly written queries that led to a monthly infrastructure bill of ®20 lakh. A single round of query optimisations reduced the server count by 40 percent, saving %8 lakh per month. For Aditya, fixing code is often more effective than adding servers. eae we & uw ig ff cergg ate 1 MMe HHHHE 93 He categorises scaling strategies based on maturity and applies tools like K6 and Artillery to support predictive decisions. His perspective on autoscaling is cautious—t is only effective when backed by data, not fear. Observability, too, must be designed carefully. Retaining logs or enabling full tracing without constraints can lead to ballooning costs. Aditya advises teams to evaluate metrics before implementation and to consider open-source tools with centralised collectors. To him, cloud cost is a shared responsibility. Developers, DevOps engineers, architects, and finance teams all influence the final bill. Making cost-awareness a part of engineering culture leads to more resilient systems and long-term sustainability. Technical Innovation has a New Voice Fresh Off the Stands, Industry Magazines by ¥ GeskyAnts AIANDHUMAN PARTNERSHIP ° Al and Human Partnership in Design Sachin M Amrit Saluja UI/UX Designer| * Technical Content Writer he presence of artificial intelligence in Case Study: Rethinking design has grown steadily, bringing both ; 1 - enthusiasm and uncertainty. Rather than Spotify’s Storyline framing it as a disruption, Sachin approached the topic as an operational shift. Teams are beginning to work differently—not through replacement, but through new forms of collaboration. In his view, the role of the designer is expanding, not shrinking. The work still demands decisions that balance user empathy, business goals, and technical structure. To ground the discussion, Sachin referenced a product design case from 2022. Spotify introduced a feature called Storyline, designed to let users explore how artists created their music. Despite initial interest, engagement dropped after launch. A focused design team was assembled to address the issue, with six weeks to iterate. They brought an Al design tool into the process, inputting internal guidelines and brand objectives. The tool responded with a range of ideas, many of which generated early optimism within the team. However, once tested, several outputs proved impractical. Some ignored technical constraints. Others conflicted with Spotify’s brand voice or user patterns. As the deadline neared, the team reevaluated the role of the Al tool. Reframing the Role of Al Rather than abandoning the experiment, the designers adjusted their workflow. The Al tool served as a concept generator. Designers identified fragments with potential and began refining them. Creative intent returned to the foreground. Constraints were re-evaluated. Technical standards and emotional clarity became guiding factors again. One designer on the team, Jordan, recognised a visual structure in the generated outputs and expanded it into a waveform visualisation. The direction aligned well with both the content and user behaviour. The updated Storyline feature showed strong results: user engagement increased by 87 percent, time spent grew by 34 percent, and feedback from artists improved. The project later received a UX design award. Human Expertise in the Workflow Sachin outlined how design continues to rely on human insight. The process involves more than aesthetic judgment. Designers evaluate choices against user behaviour, product goals, and long-term value. He noted that while Al tools offer efficiency, they do not operate within the full context of a product's lifecycle or brand narrative. He explained that real design work often involves decision points that are not visible in data. Understanding friction, anticipating how users think, and aligning systems to business intent are areas that require lived experience and team discussion. These are not elements that can be automated. * AlIAND HUMAN PARTNERSHIP Where Al Tools Contribute Despite its limits, Sachin pointed out that artificial intelligence can be valuable in key areas of the design process: ¢ Generating directional ideas during brainstorming ¢ Producing rapid layout alternatives when evaluating visual options * Accelerating early-stage drafts for storyboards or flows when time is limited He emphasised that these tools offer practical assistance, especially in fast-paced settings. They help move work forward, but they do not define the intent or direction of the final product. Design as a Practice of Judgment In closing, Sachin returned to the principle that defines his approach: good design relies on judgment, not just tools. While Al continues to evolve, its role in design will depend on how thoughtfully teams integrate it into their process. For designers, the value lies in knowing when to use Al, how to shape its outputs, and where to apply a human perspective. The future of design, as he described it, will likely include both systems—technical support that scales production and human expertise that brings clarity, relevance, and depth. GEEKWIZ ° The GeekWiz Awards celebrate those among us who have turned complex problems into opportunities Devamsh U Process Co-Ordinator Creating innovative solutions for various challenges, emphasizing automation to enhance productivity and reduce manual effort. ck Contrig or? “fo, GEEKwit Sanchi Chauhan Software Engineer II Driving the IPS-Arxara project for 1.5 years, navigating changing requirements, branding, and themes—owning everything from analysis and design to integration and testing. o/ GeskWiz WINNERS GEEKW\t Arjun V Senior UI / UX Designer Shown strong leadership across projects, managing design and tech while ensuring great team coordination—earning high client appreciation. Jahanvi Sardana Software Engineer II Onboarded Kofuku during a chaotic phase and single-handedly refactored the backend. ising Sia, Immanuel Stanley Senior Business Development Executive Onboarded three major clients with consistent dedication and high performance. GEEKw\t Ishani Sinha HRBP Collaboration within many teams * GEEKWIZ GEEkw\t Joswin J Kumar Graphic Designer Joswin’'s creative visuals and trendy design ideas were key in finalizing the client's branding, adding strong value to their visual direction. Robin Ranjan Senior Business Analyst He has been pioneer in taking Innopeak project to the next step. 58 59 THE OVERTHINK TANK ° No wrong answer honest ones. Pick one Ship fast with bugs Daily standups Design-first MVP Pay for good tools early Build for feedback Roadmap in Notion Remote forever 80-hour sprints Say no quickly Gut instinct Or this one Ship slow with polish Weekly async check-ins Code-first MVP Stretch free tiers Build for conviction Roadmap in head Back to office Strict 40-hour weeks Stay open longer Metrics-led thinking Your Answer Spot the Senior Engineer How It Works: One of these engineers is a senior. The others just have strong opinions. Can you tell who is who? O O O O (<> <> (<> (<> Engineer A Engineer B Engineer C Engineer D Types git push --force Says “Let me refactor with no fear this real quick” and calmly, fixes it silently Thinks meetings are vanishes for 3 weeks Wrote their own fine, actually Uses dark mode in Slackbot Google Docs Explains the bug Has a Notion template titled “War Room” Starts sentences with “hypothetically...” * THE OVERTHINK TANK Bots Are Skipping Steps BPM is Dead. Long Live the Agents. We overheard this at the office: Pratik: “Wait, did that startup just say their Al reroutes transactions mid-KYC?” Shreya: “Yep. Real-time compliance orchestration. BPM has left the flowchart.” Pratik: “So onboarding now changes based on user behaviour?” Shreya: “Pretty much. The form watches you while you fill it out.” Pratik: “Creepy.” Shreya: “Effective.” Pratik: “So what do we call this?” Shreya: “Agent-led ops? BPM++? Compliance jazz?” Pratik: “Let’s just say: your workflows now have a mind of their own.” Somewhere out there, a bot is skipping steps your manager thought were mandatory. And it is doing it better. MONTHLY ISSUE O6| JUNE’25 GEIK CHRO NICLES by ¥ GeokyAnts

GeekChronicles Issue 6 – GeekyAnts' AI Era, Product Wins & Community Buzz

GeekChronicles Issue 6 – GeekyAnts' AI Era, Product Wins & Community Buzz

Never Miss a Release. Get the Latest Issues on Email

Let's Innovate. Collaborate. Build. Your Product Together!