gac
Writes your git commit messages in any language, using any of 35+ LLM providers. 300+ stars, 125k+ downloads across 100+ countries. Documentation in 16 languages.
$ whoami
// Principal Software Engineer
I build agent harnesses and developer tools. Creator of gac (125k+ downloads). Principal engineer working on Code Puppy, an open-source AI coding agent.
I build the harness layer of AI coding agents: the tools, evals, and guardrails that make them trustworthy. I'm on the core team of Code Puppy, an open-source coding agent I work on full-time at Walmart Global Tech. I was hired after starting as its top outside contributor. Before that: fifteen years of data science and software engineering, and a steady habit of shipping my own tools, gac, jobhunt.coach, and kittylog.app among them.
Writes your git commit messages in any language, using any of 35+ LLM providers. 300+ stars, 125k+ downloads across 100+ countries. Documentation in 16 languages.
Open-source AI coding agent (500k+ downloads), a terminal-native alternative to Claude Code and Codex. My day job — developing both the open-source and internal versions plus the infrastructure around them.
Agentic AI career coach that runs your whole job search — oral-history interview, tailored resumes (LaTeX → PDF), interview prep — through a durable agent action loop, with an honesty guardrail on every LLM call. Also my eval workbench: 30+ custom LLM-judge harnesses for comparing models and design directions. Live, invite-gated, multi-user; built solo.
AI changelog automation for GitHub: turns commits and PRs into release notes on every tag or on a daily/weekly/monthly schedule, with audience tuning, translation, and custom domains. Live SaaS with paying customers — designed, built, and operated solo.
also: trendinginpython.com — hourly Python leaderboard (fastapi/react)
Code Puppy team. Since joining: internationalized the tool to Spanish and French, and rebuilt the browser-automation tools from brittle screenshot-and-pixel-clicking to resilient semantic locators. Got here as the #1 external contributor — 35 of 190 merged PRs (~18%) by my start date — on the strength of autocomplete improvements and deterministic safety guardrails.
Built TransplantAI, a research-stage clinical decision-support system for determining organ and tissue donor eligibility, achieving 0.99 AUROC on a held-out clinical dataset.
Energy-efficiency data systems processing billions of dollars in utility rebates; machine-learning fraud detection. Progressed from business analyst to senior data scientist over seven years.
Earlier: data analyst at Elevate Energy (2012–2014); research intern at Chicago Public Schools (2011–2012).
Education: MS Applied Statistics, DePaul University · BS, Rochester College.
Agent harnesses & skill authoring (SKILL.md) · provider-agnostic LLM orchestration · MCP · PydanticAI · RAG & ensemble retrieval · vector search · LLM-as-judge evals · deterministic safety guardrails · browser automation · usage metering & observability
Python (expert) · TypeScript · JavaScript · Node.js · SQL
FastAPI · scikit-learn · Keras · PostgreSQL · SQLite · Redis
Accessibility (WCAG 2.2 AA / WAI-ARIA) · React · Next.js · Stripe billing · GitHub Apps & webhooks · transactional email (Resend)
TDD & high-coverage testing · Playwright e2e · security engineering (encryption, rate limiting, webhook verification) · technical writing
Docker · GitHub Actions · Cloudflare Workers · Fly.io · AWS · Azure · background jobs & schedulers · object storage (S3/R2)