At AB InBev GCC, I built an in-house Deep Researcher using CrewAI, reaching 94% citation precision in RAGAS evaluation. I traced agent workflows with OpenTelemetry and Arize Phoenix, and used intent-driven prompting and memoization to manage latency.
I also built and deployed a FastAPI and Next.js application in five working days, and migrated managed Azure PostgreSQL to PostgreSQL in Kubernetes with no data loss. Earlier, at Samsung Prism, I implemented and trained a PyTorch monocular depth estimation model and reduced inference latency from 30ms to 11.9ms.
