AI & ML Engineer building systems that make it out of the notebook.
I build machine learning systems that make it out of the notebook. My work spans medical imaging, IoT agriculture, and cheminformatics — and in each case the goal was the same: a model someone can actually run, with output they can actually trust.
That means pairing a 93%-accuracy CNN with Grad-CAM++ so a radiologist can see what it looked at, or wiring live ESP32 soil sensors to a Flutter app in nine Indian languages so the farmer reading it isn't forced into English. Published patent IN202621081530 A1 (Vision-Guided Robotic Welding System) and a second patent pending (Semi-Automatic Paddy Transplanter), GATE CSE 2025 qualified, and top 38 of 193 teams at Smart India Hackathon 2025.
Three systems taken from problem statement to something runnable.
Soil-health guidance most Indian farmers can't get: live ESP32 sensor readings turned into plain-language fertiliser advice, delivered in the language they actually speak.
End-to-end geospatial intelligence platform that detects land-cover changes across Indian regions using Sentinel-2 bi-temporal imagery and a Siamese Transformer (ChangeFormerV6) — fully offline for demos, live CDSE data in production.
Predicts likely side effects straight from a molecule's structure, letting researchers screen compounds computationally before committing to lab time.
Open to internships, collaborations, and interesting problems.
