Spearheaded end-to-end MVP product strategy for a peer-to-peer AI marketplace leveraging autonomous agents for buyer-seller matching

Designed vector-based retrieval system (RAG pipeline) integrating Bedrock + Pinecone for contextual search and smart recommendations

Built AI governance and audit layer for model explainability, ensuring transparency in AI recommendations and transaction fairness

Led collaboration across design, engineering, and data science teams to deliver a serverless backend (AWS Lambda + DynamoDB) and a highly modular front-end architecture

Implemented metrics-driven experimentation to refine AI response accuracy and latency — achieving 95% query precision and <250ms response time