Software Engineer
I build software that shows its work — reliable systems, developer tools, and grounded AI.
Engineered so every claim is traceable, and missing data is never mistaken for failure.
01 Selected work
Point it at any public GitHub repository and it reports how the project is actually run — activity, PR flow, CI, documentation, hygiene — as a deterministic score where every number is traceable to the public evidence it came from.
Full Stack · AI
A local-first study system that ingests the notes and PDFs you already have, extracts the concepts worth learning, and drills them with FSRS-6 spaced repetition. Fully functional with zero AI configured.
$ processpilot status
$ processpilot top
$ processpilot explain 95707
# read-only by design — no kill,
# no stop, no exec commands
Systems · Developer Tools
A read-only macOS process monitor built from a Rust telemetry collector and a Go analysis service. It explains CPU and memory in plain language using ordinary unprivileged APIs, and has no kill switch by design.
02 How I build
01
A score you cannot audit is not worth showing. RepoSignal links every number to the public data behind it, and TraceMark refuses to save a quote without its source.
02
Scores, schedules, and verdicts come from pure, tested functions. StudyForge works fully with no AI configured — a model may explain results, never compute them.
03
Missing data is not evidence of failure. Unmeasured values stay null, Gatehouse returns UNKNOWN instead of guessing, and a rate with no sample renders as “—”, never as 0%.
03 Grounded AI
AI is most useful when it can retrieve trustworthy evidence before it reasons — and when the system around it decides what the model is allowed to touch.
04 Experience
AI engineering at SolutionStream, building client-facing product — AI-driven features, React and Next.js frontends, APIs and PostgreSQL data systems, and the QA and release validation that keeps them trustworthy.
05 Open source
Working in codebases I don’t own, on their terms. More on this
06 Writing
August 24, 2026
RepoSignal and Gatehouse both turn GitHub state into deterministic judgments — one in TypeScript and Next.js, one in C# and .NET. Building both taught me which engineering actually belongs to the stack.
August 22, 2026
When a grounded system gives a bad answer, the model is rarely the first place to look. Retrieval is where RAG systems are won, lost, and debugged.
August 19, 2026
The useful question isn't whether to use a language model. It's deciding — precisely, in the architecture — what the model is allowed to touch.
07 — Contact
I’m happiest working where reliability, tooling, and product quality meet. If that’s your team, say hello.
