Legal judgment
Federal defense work taught me that legal technology fails when it ignores confidentiality, attorney review, client dignity, and the actual burden of records-heavy practice.
Richard Zhu builds reviewable systems for law, data, and judgment-sensitive work.
I work across legal AI adoption, public defense workflow, empirical legal research, and product systems. My strongest work sits where messy records, institutional constraints, and human judgment meet: federal defense materials, appellate sentencing data, source-linked AI interpretation, sensitive telemetry, and privacy-preserving analytics.
The site is built as a proof room. The data room carries review-ready career artifacts; the project pages document technical systems; Lux and Notes keep the visual and intellectual record available outside the first-screen recruiting path.
Technical systems
My projects use Python, JavaScript, Next.js, FastAPI, Supabase, structured extraction, SQL data layers, and LLM workflows to make complex information reviewable.
Empirical method
My thesis converted federal appellate opinions into structured data and treated legal doctrine, publication status, posture, and institutional review as part of the model rather than background noise.
Field record
Lux, Notes, and the worked-places map preserve the broader record: photography, research notebooks, field context, and the evidence trail behind the systems.