AI-Powered Interview System for Unojobs: Automating Voice-Based Candidate Screening
ABOUT THE CLIENT
A hiring industry pioneer. Their product is a smart hiring platform that uses AI and automation to simplify tech recruitment. It helps companies and candidates connect faster through efficient, bias-free processes that improve hiring outcomes.
OVERVIEW
We built an AI Interview System for Unojobs that automates candidate screening through real-time voice interactions and intelligent assessments. Powered by a scalable microservices architecture, it removes manual bottlenecks, standardizes evaluations, and accelerates hiring decisions.
The result: Faster recruitment, smarter talent insights, and a consistent, scalable hiring experience across regions.
BUSINESS
REQUIREMENT
To build a fully automated AI-driven interview system capable of conducting and assessing real-time, voice-based technical interviews with dynamic question flow and comprehensive candidate evaluation.
Key Features Requested
- Conduct AI-led voice interviews
- Enable resume & job description-based question generation
- Provide real-time speech recognition and synthesis
- Implement dynamic question adaptation
- Allow pause/resume during an interview
- Detect potential cheating during interview
- Generate detailed reports for candidate performance
SOLUTION
We set out to reimagine technical hiring by building an AI-led interview system that’s fast, adaptive, and scalable. The goal was simple: automate voice-based assessments with intelligence and reliability baked in.
We didn’t just solve for automation—we delivered intelligence, empathy, and precision in every interview session.
- Modular by design – We broke the system into focused microservices for planning, Q&A, transcription, voice synthesis, and reporting.
- Real-time sync – WebSockets handled live voice and UI communication; Redis managed distributed state seamlessly.
- Smarter logic – GPT-4 and Langchain-powered contextual, resume-aware question flows that adapt in real time.
- Human-like interaction – Google Cloud Speech enabled accurate transcription, while ElevenLabs gave the AI a natural voice.
- Built-in resilience – From cheat detection to pause/resume controls, error recovery, and live report generation, the system was engineered for trust and scale.
CHALLENGES
We solved key challenges in building a real-time, reliable AI Interview System—minimizing audio latency, maintaining context across stages, syncing distributed state with Redis, and ensuring consistency under concurrent sessions. We fine-tuned STT accuracy, improved voice synthesis, and built robust async error handling for uninterrupted performance.
Managing latency during live audio processing
1
Maintaining context across interview stages and services
2
Handling distributed state using Redis
3
STT accuracy and natural voice synthesis
4
OUR APPROACH
To build a cutting-edge AI-driven interview automation system, we adopted a modular, microservices-based development strategy executed over a focused 3–4 month engagement. Our process followed an iterative rollout of features, punctuated by regular integration checkpoints to ensure alignment across services.
The execution began with decomposing the application into core microservices—planning, Q&A, transcription, text-to-speech (TTS), and reporting.
Strategic Planning & Engagement Model
Engagement Duration: Executed over a tightly scoped 3–4 month period with a feature-focused roadmap.
Development Strategy: Adopted a modular, microservices-based approach to enable flexibility, scalability, and independent deployment of key features.
Rollout Model: Employed an iterative feature rollout plan with defined integration checkpoints to align progress and validate end-to-end functionality.
Microservices Architecture & System Decomposition
Service Separation: The application was decomposed into core independent services — planning, Q&A, transcription, TTS, and reporting.
Resilient Communication: Used Redis to manage service-to-service communication and maintain application state across distributed nodes.
Session Control: Integrated real-time control capabilities (pause/resume) and error handling mechanisms across services for uninterrupted interview flows.
Real-Time Interaction Enablement
Voice & UI Sync: WebSocket was used to enable real-time synchronization between user interface and backend voice services.
Live Session Handling: Developed voice activity detection (VAD) to maintain natural conversation flow and trigger AI responses appropriately.
Context-Aware Adjustments: Enabled the system to adapt dynamically based on conversation context, allowing for question switching and session guidance.
AI & Speech Capabilities Integration
GPT-4 Interview Logic: Integrated OpenAI’s GPT-4 to power the question-generation engine with contextual intelligence.
Langchain Flows: Customized the flow of interviews using Langchain to orchestrate dynamic and domain-specific Q&A pathways.
Speech Tech Stack:
Google Cloud Speech: Provided high-accuracy real-time transcription of user responses.
ElevenLabs TTS: Delivered natural and expressive voice synthesis for the AI interviewer
Monitoring, Reporting & System Robustness
Cheat Detection: Implemented mechanisms to monitor behavioral patterns and flag anomalies during interviews.
Real-Time Reporting: Built a reporting pipeline to generate insights immediately after each session, including transcript summaries and performance metrics.
Reliability Engineering: Integrated robust retry mechanisms and failure handling to ensure consistent system uptime and user experience across edge cases.
PROJECT RESULTS
The AI Interview System built marks a significant leap in recruitment technology. By fusing real-time voice interaction with robust AI assessment, the solution transforms how companies screen and evaluate talent.
Backed by a scalable microservices architecture, it delivers automation, consistency, and intelligence—empowering hiring teams to make faster, smarter decisions.
Let’s build your E-commerce Idea together!
Connect with our team for a free discovery session.
Let's Talk
OTHER CASE STUDIES
WaxBuddy: A Mobile-First Wellness App Built for Lush to Simplify At-Home Waxing
Built WaxBuddy for Lush Wellness, integrating tutorials, schedules, and admin tools to enhance user engagement and streamline at‑home waxing.
Real-Time Monitoring Solution for Secure Storage Management | Case Study
The client aimed to develop a user-friendly app that allows facility managers and users to easily monitor the status of their storage units, ensuring safety and security. Key features such as motion detection alerts, scheduled image captures, and environmental monitoring were incorporated to provide a seamless experience for users managing their facilities.
MSAL Integration for Web and Mobile for Restaurant Leader
Streamlining application security and management for a Restaurant Leader by reducing the need for additional resources or efforts.