Not tutorial apps. Not sandboxes. Real systems — running right now, serving real people.
- 🚆 Production Vendor Platform — cut login 12× (60s → 5s) and data fetch 12× (2min → 10s). Joined as bug fixer; resolved issues faster than the existing team — manager handed full ownership across 4 railway divisions.
- 🎓 Identity Verification System — QR generation, face detection, multi-role workflow; serving 10,000+ users across 20+ institutes. Built entire backend solo.
- 🏛️ Government of India Copyright holder — Reg. L-151429/2024 for an AI image forgery detection system built with CNN + ELA.
- 📄 3 published research papers in Computer Vision and NLP.
I move fast, write clean code, and take full ownership of what I build.
| 10,000+ | 12× | 8 Live | 7 | 4 | 3 | 1 |
|---|---|---|---|---|---|---|
| Users on systems I built | Performance gain on production backend | Production apps delivered | OSS contributors led | Railway divisions deployed | Published research papers | Govt. of India copyright |
Open-source AI platform consolidating resume analysis, mock interviews, job tracking, and skill gap detection — 7 active contributors, Phase 1 backend + AI layer complete.
- Backend — 6 Spring Boot microservices: API Gateway · User · Resume · Analysis · Interview · Job Tracker; RabbitMQ async AI queue, Redis caching, Apache Tika resume parsing
- AI Layer — 5 LangGraph stateful agents: ResumeAnalysisGraph · RecruiterSimGraph · InterviewGraph · SkillGapGraph · RoadmapGraph; exposed via FastAPI with Qdrant vector search
- Leading — authored full PRD, raised 20+ structured GitHub issues, reviewing PRs, coordinating contributors across frontend, backend, and AI modules
Next.js 14 · Spring Boot 3.2 · LangGraph · FastAPI · Qdrant · PostgreSQL · Redis · RabbitMQ · Docker
Multi-tenant SaaS bug tracking with AI-powered triage — production-grade from the ground up.
- RBAC with Admin, Developer, and Tester roles — scoped data access at every layer
- AI triage — auto-assigns severity, detects semantic duplicates, generates reproduction steps via Groq LLaMA 3.3 70B
- Auth — JWT + BCrypt + Spring Security 7, containerized with Docker
- Architecture — Controller–Service–Repository + DTO, @Transactional on service layer
Java 21·Spring Boot 3·Spring Security 7·MySQL 8·React·Redux·Groq API (LLaMA 3.3 70B)·Docker
Real static analysis — not just LLM opinions.
- AST parsing via
acorn— structural analysis, not surface-level feedback - Cyclomatic complexity scoring + maintainability metrics across 10 languages
- GitHub URL mode — paste any public file URL and review it directly
- Chose Groq over OpenAI — 14,400 free req/day, same SDK, zero cost; stateless backend
Node.js·Express·React 18·Vite·Groq API (llama-3.3-70b)·acorn AST
AI system for detecting image manipulation — backed by a government copyright and 3 research papers.
- CNN + ELA (Error Level Analysis) to detect JPEG artifact manipulation with high accuracy
- REST API with heatmap overlay showing exactly where tampering occurred
- Model not publicly hosted due to size — demo available on request
Python·TensorFlow·OpenCV·Flask·CNN·ELA
| # | Title | Domain |
|---|---|---|
| 1 | AI-Based Image Forgery Detection using CNN + ELA — Govt. Copyright Registered | Computer Vision · Security |
| 2 | AI Image Enhancement using Deep Learning | Computer Vision |
| 3 | AI Mock Interview & NLP-based Feedback Analysis | NLP · EdTech |
I'm looking for SDE-1 / Backend / Full Stack roles. Immediate joiner — open to relocation and remote.
If you need someone who ships real software and takes ownership, I'd love to connect.