At ClarityX, I architected and deployed a production multi-agent platform that turns natural-language questions into location-analytics answers. Its 1,579-query Delhi NCR evaluation suite reached 0.3% hard failures.
I also architected and built the backend of ATLAS, a multi-tenant geospatial site-selection platform. I took it to production on AWS, with six geospatial scoring engines and a single JSON contract for the platform and its server-driven UI.
At MapmyIndia, I built a multilingual LLM semantic-search system integrated with place APIs, improving place-discovery accuracy by 35% against the keyword-search baseline. I also developed production computer-vision systems, including YOLO-based driver monitoring at 97%+ accuracy and ArcFace identity verification at 89%+ accuracy.
My work has also included FastAPI and Docker microservices for vision models, edge inference optimization, and QGIS tools for satellite imagery and 3D building automation. I focus on building and deploying practical AI, backend, and geospatial systems.
