Alex Lee

I’m a Senior Applied AI Engineer focused on building production-grade AI that clinicians can trust. I specialize in RAG pipelines, LLM integration, and evaluation workflows that improve factuality, retrieval precision, and retrieval-to-generation alignment.

At Tempus AI, I developed LangChain-based RAG using LLaMA 3 and OpenAI GPT-4, powering clinician-facing summarization tools. I reduced hallucinations by 60% (clinician-reviewed) by tuning chunking, embeddings, and model selection, and I improved retrieval quality for a 500K-pathology-report search experience using Pinecone hybrid keyword embedding search.

I also built secure agent foundations, including a Model Context Protocol (MCP) server to standardize compliance and traceability for agentic LLM workflows. To ship reliably, I created FastAPI async REST APIs with streaming traces, optimized caching and I/O flow to reduce average summary latency from 7.8s to 4.6s, and containerized microservices for retrieval, embedding refresh, and feedback storage.

Across my background, I bring full-stack execution and careful engineering discipline, from React/Node.js systems to typed TypeScript and CI/CD. I’ve automated model and prompt regression tests using LLM scoring plus scikit-learn precision/recall/F1 metrics, and I’ve implemented HIPAA-aligned PHI masking, IAM access layers, and audit logging to support compliance-driven AI in healthcare.