About

I like problems where the constraints are real — a video feed that has to work in bad lighting, a dataset too large to read by hand, a model output that has to be trusted before it ships anywhere near a decision. That's the kind of engineering I gravitate toward: less “can I build a demo,” more “can this survive contact with actual data and actual users.”

Right now that means an AI internship at FinArna, where I'm building VITAL — a system that reads someone's vitals off a plain video feed well enough to matter in an emergency. Outside of that I've spent the last year building agentic AI products end to end — routing, retrieval, multi-agent verification, the unglamorous plumbing that makes an LLM's output something you can actually trust.

I'd rather ship something that works for the ugly 20% of cases than something that looks good in a slide deck. That instinct is basically why I keep ending up in hackathons, and why I keep re-learning the same lesson: the hard part is never the demo, it's everything after it.

I'm studying AI & Machine Learning at BMSIT because I wanted the theory to catch up with what I was already building. Five hackathon podiums and a Best Paper Award at an IEEE conference later, it mostly has.

Open to backend, AI/ML, and full-stack engineering roles.