see ya space cowboy...
I make opaque infrastructure and AI systems legible enough to debug, automate, and trust.
I work where systems are powerful but hard to explain: auth stacks, Linux/networking boundaries, developer environments, AI coding agents, eval harnesses, and internal platforms. My usual output is the missing map — docs, tools, traces, reports, and visualizations that help teams understand what the system is doing and decide what to change next.
I help teams make complex infrastructure, developer tooling, security boundaries, and AI workflows easier to understand, reproduce, and improve.
Typical themes: opaque software and infrastructure, security boundaries, developer tooling, and AI-agent workflows — made easier to inspect, debug, and trust.
Small, bounded work tends to fit best: diagnostics, internal tools, repo or CI cleanup, observability hooks, eval or reporting passes, dashboards, or documentation that makes a system legible.
System mapping sprint. I reverse-engineer unclear infrastructure, auth flows, deployment paths, or internal tools into diagrams, runbooks, risk notes, and concrete cleanup recommendations.
Dev environment / CI sprint. Make your project easier to run, test, and deploy by cleaning up Docker, local setup, dependency drift, CI/CD pipelines, and build reproducibility issues.
AI workflow evaluation sprint. Make AI prototypes measurable by defining success criteria, capturing traces, comparing versions, and reporting what improved or regressed.
Typical starting point: $60/hr, with fixed-scope starter projects available for well-defined work.
If you are not sure how to scope something, describe the system and what feels unclear — we can pick a narrow first slice.
My Background.
Search through my back catalog.
Comparing structured retrieval, Bash, and graph-guided search policies with SearchBench.
Using Buck2 and Nix to turn repository operations into a small graph of sanctioned actions instead of a pile of debugging commands.
Getting bleeding-edge Zed running on NixOS, Asahi Linux, Wayland, and Apple Silicon because I wanted to try the new agent features before my package set caught up.
I make opaque infrastructure and AI systems legible enough to debug, automate, and trust.
I work where systems are powerful but hard to explain: auth stacks, Linux/networking boundaries, developer environments, AI coding agents, eval harnesses, and internal platforms. My usual output is the missing map — docs, tools, traces, reports, and visualizations that help teams understand what the system is doing and decide what to change next.
I help teams make complex infrastructure, developer tooling, security boundaries, and AI workflows easier to understand, reproduce, and improve.
Typical themes: opaque software and infrastructure, security boundaries, developer tooling, and AI-agent workflows — made easier to inspect, debug, and trust.
Small, bounded work tends to fit best: diagnostics, internal tools, repo or CI cleanup, observability hooks, eval or reporting passes, dashboards, or documentation that makes a system legible.
System mapping sprint. I reverse-engineer unclear infrastructure, auth flows, deployment paths, or internal tools into diagrams, runbooks, risk notes, and concrete cleanup recommendations.
Dev environment / CI sprint. Make your project easier to run, test, and deploy by cleaning up Docker, local setup, dependency drift, CI/CD pipelines, and build reproducibility issues.
AI workflow evaluation sprint. Make AI prototypes measurable by defining success criteria, capturing traces, comparing versions, and reporting what improved or regressed.
Typical starting point: $60/hr, with fixed-scope starter projects available for well-defined work.
If you are not sure how to scope something, describe the system and what feels unclear — we can pick a narrow first slice.
see ya space cowboy...
My Background.
Search through my back catalog.
Comparing structured retrieval, Bash, and graph-guided search policies with SearchBench.
Using Buck2 and Nix to turn repository operations into a small graph of sanctioned actions instead of a pile of debugging commands.
Getting bleeding-edge Zed running on NixOS, Asahi Linux, Wayland, and Apple Silicon because I wanted to try the new agent features before my package set caught up.