A builder, not an entrepreneur. These are systems I architected and shipped — some for the commons, some for the companies I've helped build. Sorted by where they live, not by how loud they were.
FOSS the whole way down. Free forever, self-hostable, no telemetry.
A zero-hallucination symbolic reasoning engine — a hallucination firewall for the LLM era. nibli KR in — a human-readable predicate-call language — first-order logic out; every answer derived, never guessed, with a proof trace on every conclusion.
A single-binary Rust automation framework for acceptance testing and RPA — plain-text human-readable syntax, extensible via Rust, Python, and JavaScript.
An agentic English→Lojban translator that produces *verified* Lojban: an LLM drafts a translation, three real Lojban compilers reject anything invalid, and the exact compiler error is fed back to the model until the output passes every gate. Runs entirely in your browser — bring your own LLM key, no server, and the grammar/semantics gates run on-device with zero network.
A pure-Rust, rule-based Lojban speech synthesizer that runs entirely in the browser — Lojban text in, natural speech out, no server and no network after load. Ships a tuning console for crowdsourcing the voice; every config replays bit-identically on the native CLI.
This website — a Zola static site that ships its own AI: two LoRA fine-tuned twins and the real nibli engine, all running in your tab via Rust→WASM. The site is its own demo, and the source is open — training pipeline included.
The technical details a programmer should consider before making a web app public — security, performance, operations. My most-starred repository.
A collection of resources to learn from and become a self-taught awesome programmer — the path I actually took, written down for the next dropout.
A Python CLI to block and unblock websites straight from /etc/hosts — ad lists, trackers, malware domains, or just the feeds at 2am. Simple and easily extendable.
A curated map of the Rust machine-learning ecosystem — frameworks, kernels, and LLM runtimes. Curation is also building: someone has to keep the index honest.
Systems I architected as Software Architect and, before that, VP of Engineering. Some closed, all real.
Technical Architect, since May 2026 — backend engineering for niche tech initiatives.
Two tenures. 2014–17: architected the mobile-security scanning platform from scratch, built the team 1→5, ran an iOS/Android device farm. 2022–23: a Rust microservice generating CycloneDX SBOMs from mobile binaries — no source required.
Principal Architect, Distributed Systems/Rust (2023–25). The first MVP of a distributed WebRTC media server in Rust, plus the entire DevOps lifecycle with Pulumi infrastructure-as-code.
CTO for five years (2018–23). A company with a non-profit ethos: RobotFramework QA services, and teaching people with zero tech background to start QA-automation careers — unpaid, on purpose.
Principal Architect, Platform (2023–26) — on the airline-retailing platform I originally built from scratch at Reckonsys. Highlight: rewriting the NDC adapter from Python to Rust.
VP of Engineering (2019–22), ~70% hands-on by choice. Django/GraphQL/React for startups, teams of 2–10; open-sourced 'bigga' and the cookiecutter templates everything was scaffolded from.
An iOS mindfulness app (2012, at Tataatsu) — meditation and attention training, with an API server for scores and leaderboards. Designed, developed, and deployed as a one-person team.
“One must imagine the cluster happy.”
have a hard systems problem worth building?
