I build AI systems that take real actions: multi-agent pipelines that make decisions, explain why they made them, and send them through human approvals before anything is executed. I work mostly in Python, FastAPI and PostgreSQL, from the ML model through to a Dockerised, tested deployment.
- 🎓 B.E. Computer Engineering graduate, SPPU — First Class with Distinction (CGPA 8.34)
- 📄 Published: Smart Industrial Safety Wearable (AI + IoT), IJSART, Nov 2025
- 🚀 Shipped: 3 Android apps on Google Play + web on one Python REST API (bestfoodservices.in)
- 🔭 Now building: open-source agentic AI and RAG tools for Python developers
- 💼 Open to: Agentic AI · GenAI / LLM · AI Automation Engineer roles
SmartStock RPA — agentic AI + workflow automation for inventory & purchasing
Eight AI agents run every morning to forecast demand, calculate reorder points and order quantities, rank suppliers and flag anomalies. Their decisions then go through an audited approval and purchase-order workflow, from request to supplier to goods receipt to three-way invoice match, so a person stays in control.
| Results (90-day back-test) | forecast MAPE 40.3 % → 35.2 % · stock-outs −74 % · expired stock −92 % |
| Engineering | 162 FastAPI endpoints · 49 PostgreSQL tables · 178 backend tests (90 % core coverage) · Playwright e2e |
| Stack | FastAPI · SQLAlchemy · Celery · scikit-learn · React + TypeScript · Docker Compose |
| Repository | What it does | Highlights |
|---|---|---|
| agentgate | Multi-agent orchestration with human approval gates | Rule or LLM planner · typed tool calling · parallel steps · retries/timeouts · SQLite memory · traces & metrics · FastAPI · 23 tests |
| ragforge | Hybrid-search RAG for PDFs and docs with cited answers | Semantic chunking · BM25 + vectors (RRF) · reranking · query rewriting · evaluation pipeline · works offline · 17 tests |
| repolens | AI software engineering assistant for Python repos | Repo analysis · 18 bug/security rules · verified test generation · docs generation · PR review & risk scoring · 47 tests |
Tested on Python 3.10–3.13 in CI · MIT licensed · open roadmap issues welcome contributions
| Project | What it does | Stack |
|---|---|---|
| RAG Framework for Technical Support | Q&A over your own docs: FAISS retrieval → cross-encoder rerank → grounded answers with citations; voice in/out | LangChain · FAISS · FastAPI · RAGAS |
| Multimodal Fake News Detection | Fuses text, propagation-graph and image signals with an explainable dashboard and feedback retraining | PyTorch (BERT + GCN + CNN) · FastAPI |
| Industrial Safety Wearable 📄 | Wearable IoT node + YOLOv10 PPE detection with emergency escalation | ESP32 · YOLOv10 · Flask · React · MongoDB |
| Blockchain Credential Verification | Tamper-evident degrees with W3C DIDs/VCs, verifiable offline | Solidity · Hardhat · React |
| Best Food Services Platform 🟢 live | 3 Play Store apps + web on one Python REST API with JWT / RBAC | Python · REST · Android |
AI / LLM: agent orchestration · tool / function calling · RAG (hybrid search, reranking, citations) · LLM evaluation (RAGAS) · forecasting · anomaly detection · computer vision (YOLO)
- Full-Stack & AI Developer — Best Food Services (freelance) · Jun 2026 – present
- Engineering Trainee, Python Full-Stack — ProAzure Software Solutions · Dec 2024 – Feb 2025
- Cybersecurity Intern, VAPT & SecOps — Redynox · 2024
Some showcase repositories contain architecture, results and screenshots only. Their source is private and available on request for hiring or academic review.