SkillSync AI
AI-powered hiring and assessment platform — backend architecture, GPT-driven resume parsing, real-time transcription, and video analysis for structured candidate evaluation.
What it is
SkillSync AI is an intelligent recruitment platform that automates candidate assessment through AI-driven resume analysis, live video interviews with real-time transcription, and structured scoring. I owned the complete backend — from auth and session handling to GPU-accelerated video processing pipelines.
What I built
I designed and implemented the full backend architecture for this platform. The system handles authentication with cookie-based sessions, role-based access control, secure AWS S3 file uploads, and production-grade observability.
AI and processing pipelines:
- GPT-powered resume parser that generates candidate summaries, skill-match scores, and role-fit insights in real time
- Gladia streaming transcription with async multithreaded processing to evaluate responses, assess core-value alignment, and score answers
- Haar Cascade facial analysis for detecting engagement signals, emotion patterns, and face count during video assessments
Platform systems:
- Custom job-positioning module connecting assessments, skills, interviewers, and hiring workflows into a structured evaluation pipeline
- Hybrid data layer — PostgreSQL with Sequelize for relational data, MongoDB for nested documents like transcripts, assessment responses, and analytics
- Detailed access logs, application logs, and error tracking for debugging and API usage monitoring
Frontend integration:
- Connected the full UI to backend APIs including WebRTC-based video recording, real-time audio extraction, and an embedded live code editor for technical assessments
Tech stack
- Node.js — API server and processing pipelines
- PostgreSQL + Sequelize — structured relational data
- MongoDB — nested documents, transcripts, analytics
- OpenAI GPT — resume parsing, ATS optimization, scoring
- Gladia API — streaming speech-to-text transcription
- AWS S3 — secure file storage and uploads
- WebRTC — real-time video and audio capture
- Haar Cascade (OpenCV) — facial analysis and engagement detection