I build systems that make their own reasoning inspectable.
- 🛰 Remote sensing — recovering terrain from lunar craters that have been dark for two billion years
- 🤖 Multi-agent systems — agents that argue, where the disagreement itself is the auditable artifact
- 🔗 Verifiable state — hash chains written from scratch, with tampering demonstrated live
- 🧪 Developer tooling — a code reviewer graded by the repository's own future
A theme runs through all of it: the output has to carry its own evidence. Every project below ships the numbers that justify it — and reports the ones that don't.
PSR-Net · SPARC · CarbonLedger — PSR-Net pipeline · SPARC evidence sheet · FancePro ruling
Languages
AI & scientific computing
Frontend
Backend & tooling
Everything here is open source and built to be read. Every repository carries a contributing guide, issue templates, and a pull-request checklist written against that project's actual constraints. SPARC and PSR-Net have the most surface area for contributors — methodology docs, an API contract, and a reproducible evaluation script.
