AI-native product engineer · founding @ Auptim · Spokane → remote
I find the gap between the official model and reality, then build the system that closes it. I take ambiguity and AI-generated chaos and turn them into secure production. AI as leverage; judgment intact.
Early engineering roles at AI-native startups, remote. Founding engineer, second hire, MTS. I want a role where one person still owns whole systems: a real product with real users, a hard, ambiguous problem, and a quality bar that doesn’t move. My best work happens in the mess. Ambiguity, AI-generated chaos, a half-built prototype that has to become something you can ship. Implementing already-solved problems bores me.
Agents write the code I ship. They do the typing; I own the architecture, the taste, and the verification. If you’re hiring an early engineer who can move at AI speed without handing over judgment, get in touch.
Live products, open source, and tools built to solve a real problem in front of me.
AI-native project management with an AI discovery → SRD → backlog pipeline. Inherited an unstable AI-generated prototype and reimplemented every feature myself, page by page. Re-architected the agent loop from one-shot JSON into tool-driven and validated, closed IDOR and permission gaps, added 700+ tests, and de-slopped the whole codebase into secure production.
Repository intelligence for AI coding agents. Turns a codebase into structured signals — symbols, dependencies, blast radius, duplication, complexity — and gates diffs before they ship. Required in my own daily agent loop; my agents don't ship code without it.
A complete Three.js game I conceived and built as the studio's engineer on a two-person team, shipped cross-platform to web, desktop, and mobile from one codebase. Real physics, leaderboards, accounts, and a custom build pipeline that fans out to Electron and Capacitor.
42K
downloads · 9 models
Custom ROCm FP4 4-bit quants for AMD Strix Halo: Qwen3, Deckard, and Nex variants with MTP self-speculation. Contributor to a ROCm quant format for llama.cpp.
Two products I built start to finish, then set down. Both reached robust, fully working prototypes — real auth, payments, and data models — before they were shelved ahead of launch. Ready to pull back off the shelf whenever they're worth pursuing.
Stable-management platform: horse tracking, incident reporting, subscription billing. Inherited a rough build and took it to a feature-rich, fully working state — full RLS, payments, and a secure multi-tenant model.
booking · discovery · payments
A booking platform, architected and built solo — row-level security, 100+ migrations, Stripe payments, and LangChain/OpenAI agentic flows for discovery and routing. Reached a complete, working build.
Five ops tools, built unprompted. Four are still in daily use at Amazon; one runs nationwide. One quantified a year of sortation-machine jams in roughly a minute. Another recommended optimal stow locations with per-zone PDF guides; it still runs at my old FC, and its code was later pulled into a directed-stow tool used at FCs nationwide. The rest were data-analysis tools that augmented operational dashboards and generated slide decks for meetings. Taught myself to code to ship them.
Each step took on more of the whole.
Primary author across three AI products. Took over two unstable AI-generated codebases — rebuilt one into secure production, brought the other into a feature-rich pre-production state — and built the third entirely solo.
An independent game studio. I'm the engineer on a two-person team: I conceived and built Synth Runner and shipped it cross-platform to web, desktop, and mobile, with the studio's founder shaping its direction throughout. Now building Mid Manager, a high-fidelity warehouse-inbound simulation.
Full client sites across several industries: Poorman Legal (law), CameraWorks Media (real estate), and Counseling 2 Thrive (counseling).
Built tools nobody assigned, shipped to fulfillment centers across North America. Taught myself to code on the job.