Royce Feng

I’m a results-driven Staff AI/ML and software engineer with 12+ years building production-grade data and AI systems in healthcare. I specialize in turning complex AI capabilities into dependable, user-facing software—especially in regulated environments where reliability and safety matter.

At Tempus AI, I led architecture and productionization of a multimodal Clinical AI Platform delivering personalized oncology insights and clinical trial matching at scale. I designed and productionized LLM-powered RAG pipelines over a proprietary multimodal knowledge base, improving physician adoption by 35% while reducing hallucinations via grounded retrieval and citation mechanisms. I also built lakehouse-based ingestion and feature computation pipelines that cut feature generation latency by 60%, and I implemented HIPAA-regulated MLOps and observability (monitoring, drift detection, explainability, evaluation dashboards) to achieve 99.5% system uptime.

Previously, at Databricks, I built enterprise Feature Store and core MLOps capabilities adopted across 20+ customer deployments, reducing training iteration time from weeks to days and improving deployment reliability by 45%. Earlier at Optum, I architected and scaled clinical and claims data pipelines, improving pipeline runtime by 55% and raising reliability to 99.8% through orchestration modernization and monitoring with Airflow.