Ziyu Han

I am a Senior AI Engineer with nine years of experience designing and shipping production-grade AI systems at companies including Airbnb, Supabase, and Indeed. I specialize in LLM-powered systems, RAG pipelines, vector search, and multi-agent workflows that turn ambiguous business processes into auditable automation.

My background blends strong backend engineering with AI systems design: I build low-latency inference services, retrieval-augmented generation pipelines, and hybrid vector/PostgreSQL retrieval architectures. I have owned AI features end-to-end from data ingestion and feature computation to model serving, observability, and rollback.

I have a proven track record improving real-time inference reliability (≈40% improvement) through caching, adaptive throttling, batching, and robust fallback strategies. I’ve developed LLM evaluation infrastructure including prompt regression tests, accuracy and consistency checks, safety filters, and drift monitoring to enable safe iteration in production.

I partner closely with Policy, Trust & Safety, and Platform teams to convert human workflows into deterministic, auditable AI-driven automation suitable for high-scale, high-risk environments. I prioritize reliability, safety, and measurable impact when deploying AI features used by millions.