At Lyft, I architect reusable LangGraph orchestration components for the Enterprise Multi-Agent AI Platform & AgentOps, supporting seven production agents across approximately 270,000 monthly rider and driver interactions. I also standardized MCP tool contracts, reducing failed tool executions by 28%.
I led integrations for the Agentic Customer Operations & Case Resolution Platform with Claude through Amazon Bedrock and authorized support APIs, contributing to an 87% reduction in average resolution time. I established LangSmith evaluation pipelines that reduced production support-agent hallucinations by 20%.
Before my current work in applied AI, I built ML infrastructure and real-time feature platforms at Lyft, including serving and streaming improvements that reduced feature lookup latency and freshness lag. At Airbnb, I developed guest and host workflows and reusable React components for marketplace experiences.

