San Francisco, California, United States
Architected and led the development of a high-fidelity AI "Eval" system, allowing users to simulate, test, and score complex agentic workflows across varied LLM models and prompts in a side-effect-free sandbox environment.
Engineered responsive viewport-focusing and coordinate-mapping logic for an interactive HTML canvas state-machine, implementing algorithms for auto-fitting, dynamic zoom management, and intelligent element centering.
Designed a transition from a legacy code-generated API SDK to a robustly typed API Contract service, reducing TypeScript compilation times by several seconds and significantly improving system-wide type safety and developer productivity.
Developed and maintained a distributed AI state-machine architecture using the Temporal Workflow Engine and a React Relay/GraphQL front-end, managed via Kubernetes on Google Cloud Platform.