Matthew Liu

I’m a Senior AI Engineer focused on agentic AI systems, multi-agent orchestration, and large-scale LLM platforms. I build and scale production AI that turns unstructured data into automated workflows and actionable insights—prioritizing reliability, performance, and production readiness.

At Notion, I spearheaded NotionAI meeting notes, creating a production-grade agentic AI pipeline (capture → transcript → reasoning → action generation) with GPT-4/4o, Whisper, and streaming. I also designed and scaled multi-agent orchestration with LangGraph and LlamaIndex, improving task completion rates by 40%, and re-architected real-time inference to cut median latency from 3.8s to 1s under production load.

I’ve strengthened quality and observability end-to-end—building RAG pipelines (pgvector, hybrid BM25 + embeddings) that improved factual accuracy by 35%, implementing an LLM evaluation framework that reduced hallucinations by 28%, and adding agent observability with OpenTelemetry to reduce debugging time by 50%. Previously at CourseHero and earlier core platform roles, I engineered scalable AI microservices, TB-scale ETL/data pipelines, and search/recommendation systems that supported multi-million user platforms.