On PatentEase, I own the multi-agent orchestrator on a four-person team and built a verification layer that checks generated references against source records, removes unsubstantiated ones, and logs why. The platform combines six agents with hybrid vector and graph retrieval, and its CI includes 200+ tests.
I’ve also shipped HotShot, a live polling platform with results that update as votes arrive, and built a document-to-structured-data pipeline using Gemini API and OAuth-authorised publishing. As a Research Intern at CoDMAV - Centre of Data Modelling, Analytics and Visualization, PES University, I benchmarked deep learning backbones and built a dashboard; I was first author on the resulting peer-reviewed paper at IEEE INDICON 2025. I’m currently a Teaching Assistant at PES University, teaching and assessing courses across six sections.

