Building intelligent systems — from edge AI on a Raspberry Pi to production ML pipelines.
portfolio · linkedin · leetcode · x · maillanguages: [python, c++, typescript, javascript, c]
ml: [pytorch, tensorflow, scikit-learn, opencv, onnx-runtime]
llm: [langchain, faiss, huggingface, ggml]
web: [next.js, react, fastapi, node, express, tailwind]
infra: [docker, aws, azure, gcp, postgresql, supabase]| devtrack | Open-source developer productivity dashboard — GitHub stats, PR metrics, streaks & goals | typescript |
| contextrot | Find where your coding agent starts degrading — local context-rot analytics, zero-config via uvx contextrot | python |
| sentinel | Fully offline AI wellbeing chatbot — private by architecture, no network required | python |
| keeptrack | Chrome extension that captures download intent at download time — zero telemetry, fully local | javascript |
| omnivoice | Ultra-low latency real-time voice bot — OmniVoice TTS + Pipecat + Groq + faster-whisper | python |
| lunar-ice | Lunar ice detection for ISRO Bharatiya Antariksh Hackathon 2026 (PS-8) | python |
Robotic Arm Fault Detection using CatBoost Classifier — IEEE Sensors Letters, 2026
97.20% accuracy · F1 0.9718 · outperformed SVM, LR, NB and QDA baselines
paper: ieeexplore.ieee.org/document/11359621 · code: implementation




