At Meta, I develop Python-based LLM agent workflows for commerce, connecting conversational assistance with catalog search and business tools. I implemented LangGraph orchestration with conversation state and error recovery, and integrated embedding-based retrieval to ground responses in catalog content.
I also designed benchmark evaluations for tool-use accuracy, task completion, and unsupported claims, and added execution tracing and pytest regression coverage. Previously, at NetEase, I developed Python backend services, data pipelines, and REST APIs for an e-commerce analytics application.
