Senior full-stack & AI engineer · remote, Colorado Springs
I can’t leave a slow build alone, and I lean on AI to ship more than one person probably should.
Open-source tooling for Claude Code, a ~300,000-line platform rebuilt solo in about four months, and consulting that gets fixes to a live preview the same day.
The marquee · open source
Claude Code Infrastructure Showcase
Claude Code skills sit dormant until you remember to invoke them. This makes them activate the moment they’re relevant instead.
Specimen · CCIS
skill auto-activation
prompt + files -> skill-rules.json -> regex · AI -> skill
9.7k stars MIT4 skills / 9 hooks / 8 agents
GitHub stars, as of Jun 2026
- multi-provider AI · regex fallback Gemini / OpenAI / Anthropic / local: engages only when regex is unsure
- SQLite vector search over sessions pulls context from past transcripts and dev docs
- two-try blocking PreToolUse guard holds an edit until the right skill is consulted
- self-verifying install · 8 checks a setup wizard that proves itself before declaring success
A hook reads each prompt and the files in play, scores them against a rules config, and surfaces the right skill on its own. Matching runs on cheap regex first; the optional model layer only engages when the regex is unsure, so it costs nothing to run by default and degrades gracefully when no API key is present.
Underneath are the parts that make it hold up in a real codebase: a guard that blocks an edit until the relevant skill has actually been consulted, SQLite-backed vector search over past sessions, and progressively-disclosed skill docs kept under a size budget so they never blow the context window. It’s a reference library, not a service: the patterns and the working code behind a system you install in about fifteen minutes.
It grew out of six months of daily use on a production TypeScript codebase. The public write-up and the repository have since drawn a steady stream of inbound interest from engineers adopting the same patterns.
Selected work · described, not named
A legacy lifecycle & project-tracking portal, rebuilt from zero.
- ~300K lines
- ~4.5 months
- solo
- React 19 / TS
- Node / Prisma
One of the systems I’m proudest of is one I can’t show you. It’s a large lifecycle and project-tracking portal (years of accreted features across many modules) that had grown brittle enough that teams had started working around it. I proposed rebuilding it from zero, secured the buy-in, and shipped the full replacement solo in about four and a half months.
The result is roughly 300,000 lines: a React 19 and TypeScript front end on a typed Node and Express back end, a redesigned relational schema with its history migrated forward, and a workflow engine driven by JSON definitions and an event queue, so changing a process is configuration, not a code change. Every meaningful action is audited end to end, notifications are unified across in-app and email, and the whole system ships through containerized CI with per-branch preview environments.
I’ve kept this deliberately vague: no client, no screenshots, no schemas. If we talk, I can walk through the architecture and the decisions in the right setting.
Consulting
I take a few client engagements a year.
- ~44K lines
- 48 PRs
- first 10 weeks
- Next.js 16
- Supabase / Prisma
For The Bloom and the Bull (a compliance platform for a precious-metals dealer), I’m the sole engineer, end to end: requirements, build, and the client relationship. It’s a Next.js 16 and Supabase application with real regulatory weight: signed transactions, statutory hold tracking, live spot pricing, generated receipts.
The part the client feels most is the loop. A bug or an idea posted in their Discord becomes a triaged item, a fix, and a deployed preview they can click, usually the same day. They report from where they already work; I ship to a preview they can try before it’s real. That tightened feedback loop is what turns a vendor into a partner.
See it live -> thebloomandthebull.com shared with the client’s permission
Writing & talks
Notes on building with Claude Code.
About
I’m a full-stack and AI engineer with seven years building production systems, now splitting my time between senior engineering work and a small consulting practice. I work remotely from Colorado Springs.
I like the unglamorous parts: the rebuild nobody volunteered for, the tooling that makes a team faster, the audit trail that makes a system trustworthy. I move quickly because I lean hard on AI-accelerated workflows, and I build the infrastructure that makes that leverage real for other engineers.
Outside the editor I keep carnivorous plants and jumping spiders; my projects tend to take their names from that world. Same instincts, really: patience, close observation, and an interest in how small systems catch exactly what they need.
Let’s talk.
The fastest way to reach me is email. If you’re hiring for a senior full-stack or AI role, or you have a system that needs rebuilding, I’d like to hear about it.