Project case study / 01
HireSense
A secure, role-aware hiring system with AI assessment and live proctoring.
Award-winning working platform · Three-person team · backend ownership
At a glance / inspectable evidence
Project evidence
- Role-aware workflows
- 07
- Person team
- 03
- CYGNUSA Elite Hire
- 1st
Administrative, hiring, assessment, and candidate journeys
Akshat owned backend and AI engineering
Working platform recognised at SRMIST
01 / Overview
- Role
- Backend and AI engineering
- Stack
- Python · Flask · React 18 · PostgreSQL · OpenAI · WebRTC · Socket.IO
- Recognition
- 1st place · CYGNUSA Elite Hire
A full-stack recruitment platform spanning résumé intelligence, structured assessment, a browser coding environment, analytics, and real-time WebRTC proctoring.
02 / Context & problem
The problem
Hiring workflows often split candidate screening, assessment, proctoring, and decision support across disconnected tools, making context and accountability difficult to preserve.
The response
HireSense brings the workflow into one role-aware platform, combining document parsing, AI-generated assessments, a coding environment, live supervision, and evidence-backed review.
03 / System path
Follow the work from input to outcome.
- 01
Résumé intake
- 02
AI parsing
- 03
Role-aware assessment
- 04
Live proctoring
- 05
Evidence review
04 / Contribution
My contribution.
- 01
Built and hardened Flask APIs for authentication, RBAC, recruitment, assessment, and proctoring workflows
- 02
Integrated AI résumé parsing, candidate-match scoring, and question-generation paths with reviewable outputs
- 03
Implemented backend session boundaries for WebRTC and Socket.IO proctoring evidence
05 / Engineering pressure
The hard parts.
- 01Coordinating seven role-aware workflows without leaking privileged actions
- 02Keeping live proctoring sessions isolated and resilient
- 03Turning unstructured résumés into useful, reviewable assessment inputs
06 / Outcome & reflection
What came out of it.
The three-person team secured first place and a ₹10,000 prize in the CYGNUSA Elite Hire challenge at SRMIST.
Lessons carried forward
- AI output needs visible review paths, not just confident generation.
- Authorization is a product architecture concern, not a final middleware task.
- Real-time features become manageable when session boundaries are explicit.