π Chandigarh, India Β· Python Full Stack Developer Intern Β· Open to SDE roles
| Project | What it does | Proof it works |
|---|---|---|
| ποΈ Pix β offline voice AI for macOS | Say "Hey Pix," ask about anything on screen. Wake word β on-device Whisper β OCR β PII redaction before any model sees the frame β vision LLM β streaming TTS. Zero cloud calls. | 109/109 tests Β· STT 98% word accuracy Β· tool router 31/31 Β· packaged releases |
| π Trade Bot | LSTM signal bot placing live paper trades via Alpaca, controlled from a FastAPI server. Built a directional-accuracy evaluator that exposes why the standard 97% metric is meaningless. | Real numbers, published: 52% overall Β· 57.8% on BUY Β· 3 documented fixes |
| π Motor Volt | Django car marketplace β OTP email auth (Brevo), multi-image listings, ML price predictor, admin dashboard. Now hardening: pytest + CI + Postgres migration. | Live on Render |
| π§ Fairy World | Kids' subscription platform β Stripe payments, WebSockets, PostgreSQL, Daphne. | Live on Render |
| Real-time travel-planning chatbot β Django Channels, hybrid fuzzy/TF-IDF location search. | Test suite included | |
| π₯οΈ Remote Desk | Remote desktop built from scratch β screen streaming + remote input over raw TCP sockets. | Pure Python, no frameworks |
π§ Hardening motor-volt β pytest suite, GitHub Actions CI, SQLite β PostgreSQL π± Contributing to open source β first PRs to libraries I use daily π§ Grinding DSA β NeetCode 150, pattern-tracked with retry-sheet discipline
π¬ Ask me about Django, DRF, sockets, or why 97% model accuracy can be a lie

