I built Curio, a live AI learning workspace for chatting with PDFs, websites, YouTube, and notes. Its 70-question evaluation reached 1.00 Hit@5 and 0.98 faithfulness, with 100% correct refusals and zero false refusals.
With Hive, I built a multi-agent research system where a critic checks findings against evidence and reports cite only undisputed findings. Its live agent graph, tool-call inspector, and run replay make the research process inspectable.
As an Independent AI Engineer, I build and ship LLM products with evaluations, tracing, and cost controls. I also won 10th place among 3,000+ participants in the OpenAI x NamasteDev Codex Hackathon, where I built CommentSense solo.
At Zapygo Energy Pvt. Ltd., I built the working MVPs for the Customer app, Driver app, and Admin dashboard, including Node.js/Express and PostgreSQL APIs, Google Maps integrations, Razorpay payments, and React interfaces. Earlier, at SkillDzire, I built a water-potability classifier and compared seven models.
