Public Defense AI Workflow Pilot
A concise, risk-aware proposal for attorney-supervised AI workflows around record triage, mitigation support, and source-linked review.
Read the pilot proposalI build legal AI, empirical legal-data, and workflow systems where source control, confidentiality, and human judgment matter. The work spans federal defense, appellate sentencing research, bilingual AI knowledge tools, geospatial telemetry, and privacy-preserving analytics.
A concise, risk-aware proposal for attorney-supervised AI workflows around record triage, mitigation support, and source-linked review.
Read the pilot proposalA one-page summary of a large-corpus appellate sentencing project: LLM extraction, validation, Bayesian modeling, and legal insight.
Read the thesis briefAcross the artifacts, the recurring pattern is reviewable systems built around messy records, institutional constraints, and judgment-sensitive workflows.
Tool evaluation, confidentiality framing, attorney training, pilot design, and adoption documentation for legal environments where trust is the bottleneck.
Structured extraction from federal opinions, model-output validation, and statistical analysis that keeps legal doctrine and institutional context visible.
Full-stack systems using Next.js, FastAPI, Supabase, Python pipelines, Deck.gl, and LLM workflows to turn private or complex data into usable interfaces.
Federal defense work that required confidentiality, record discipline, client-centered thinking, and practical support under time pressure.
The data room contains review-ready PDFs and readable web versions of the most relevant proof artifacts: legal tech resume, analyst resume, public defense AI pilot, thesis brief, and a redacted federal defense project sheet.
Review artifactsThe project portfolio adds substantiated technical systems: I Ching Studio, Reizentraj, wxdecipher, Vitals, and the Via Sacra archive.
Explore systems