Projects

Building production AI and agentic systems on top of frontier models from OpenAI, Anthropic, and Microsoft AI. The hard part isn't calling a model — it's making agents, retrieval, and tools reliable and fast at millions of requests per day.

When Copilot's signals regress, millions of users get worse answers. I owned the distributed infrastructure and validation layer processing 10B+ events per day — the systems that catch problems before they reach people.

Moving Microsoft 365 data at massive scale into Azure for analytics. Distributed data pipelines where correctness, throughput, and tenant isolation are non-negotiable — the systems foundation the rest of my work is built on.

Logistics misallocation affects real businesses and real shipments. Built the simulation and ML pipeline infrastructure that made RL-driven optimization actually work in production.

Intensive care units operate under extreme resource constraints — when beds and equipment are misallocated, patient outcomes suffer. Applied machine learning to model and optimize ICU resource allocation.