Skip to Content

William Du

Founding ML Engineer

William leads agentic evaluation and performance at Legion Intelligence. William dreams about auto-improving agent systems, ML system architectures and how to incorporate AI into user workflows seamlessly. Before Legion, William led applied machine learning teams and the technical roadmap for the natural-language model training platform at Primer. He’s most of his time thinking about how ML and AI can be deeply integrated into a product and how it can drastically improve day to day lives of the labor force.‍

No items found.
No items found.
11 minute read

Why Retrieval Agents Fail: It’s Not Just the Model

Retrieval-focused agents fail for reasons beyond model capability. Performance is bounded by three independent constraints: the model, the retrieval stack, and prompt guidance. Evaluation finds the real bottleneck.