I'm an engineering student at New Horizon Institute of Technology in Thane, and I spend most of my time building AI stuff, mostly around LLMs, agents, and computer vision. I like the messy, end-to-end kind of work: from model to deploy to the thing actually running in production.
Right now I'm a Software Engineer at Qoneqt, shipping AI agents that handle 20,000+ verifications a day and personal agents for about 250,000 users. Before that, as an AI Engineer on the same team, I rebuilt their Aadhaar verification pipeline and pushed accuracy from 48% up to 95%.
I also did Google Summer of Code with SAT Montreal, porting the Puara Gestures library to the Avendish runtime so musicians could use it live on stage without latency getting in the way.
Earlier, at Stamp 'IT, I squeezed a YOLO detection pipeline to 30+ FPS at 720p on a Raspberry Pi and built a Qt-based CAD tool that cut their setup time in half. Good reminder that making things fast on small hardware is a skill of its own.
/ Skills
Programming
AI / ML
Computer Vision
Backend & Infra
Tools
/ Experience
- Rebuilt the Aadhaar authentication pipeline from the ground up. Accuracy jumped from 48% to 95% and manual review time dropped by about 80%.
- Put together an LLM-powered mobile app testing agent with Ollama + Maestro that catches 3× more critical bugs before release.
- Worked with SAT Montreal on porting the Puara Gestures library to the Avendish runtime so musicians and live performers could use it on stage without latency getting in the way.
Software Engineer
Stamp 'IT · Thane
- Squeezed a YOLO detection pipeline down to run at 30+ FPS on a Raspberry Pi at 720p. Mostly quantization, some memory tricks, and a lot of profiling.
- Built a Qt-based desktop CAD tool with real-time camera calibration and automated geometric optimization that cut their deployment setup time in half.