The Work
You'll build and extend internal web applications used daily by a large corporate physical-security operation: live and recorded video from thousands of cameras, credential and access management, incident and audit tooling. Real production software with real operators depending on it. The portfolio spans a modern React stack on one system and a lean vanilla-JS video platform on another — plus the APIs, databases, and Windows infrastructure underneath both.
How We Work
- AI-first. We plan with LLMs, generate with LLMs, and hold the human bar at review, verification, and judgment. Agentic tooling isn't an experiment here — it's how the work gets done.
- Fail fast. A working prototype against the real system beats a design meeting. We build the quick version, learn from it, throw away what doesn't survive contact, and keep what does.
- Velocity over ceremony. Small team, short cycles, frequent releases to real users. Ideas go from conversation to running in production in days, not quarters — and the feedback loop with the people using it is direct.
Who We're Looking For
Someone genuinely obsessed with AI — not as a resume keyword, but as the thing you spend your nights and weekends on anyway.
- You build with AI agents daily. Claude Code, Cursor, Codex, or your own scaffolding — you've moved past chat-and-paste into agentic workflows, and you can articulate what you delegate, what you never delegate, and how you catch the model being confidently wrong.
- You've set up agents yourself. Wired up tool use, MCP servers, automations, pipelines — anything where a model does real work unattended and you had to make it reliable.
- You tinker. A homelab, self-hosted services, local models on your own GPU, scripts automating your own life — we want theperson who builds things nobody assigned them.
- You have a portfolio. Personal projects, a GitHub that shows activity, a blog, demos — finished, working things you can point at. This matters more to us than years of experience, and early-career candidates are explicitly welcome.
- You're self-driven. Small team, lots of autonomy. You'll get problems, not tickets — and you'll be expected to chase them down, form your own hypotheses, and ship without someone standing over you.
- You follow the field. If Karpathy's "Software 2.0," spec-driven development, or the latest agent architectures are things you've actually practiced rather than just read about, you'll fit right in.
The Stack
Core — you should be able to get productive here quickly:
- JavaScript/TypeScript across the full stack
- Node.js APIs (Express, REST, JWT/session auth)
- Modern React (Vite, component libraries like MUI, React Query, Tailwind)
- SQL — schema design and queries; we run both SQL Server and PostgreSQL
Bonus — any of these move you up the list, none are required:
- Windows Server, IIS, PowerShell
- Active Directory / LDAP, enterprise SSO
- Video and streaming: RTSP, ffmpeg, H.264, MSE/WebCodecs
- ML in practice — we already run in-browser face detection, and there's plenty of room for more
- Physical security domain experience (VMS, access control) — learnable; the mindset above is what we're hiring for
How To Apply
Skip the cover letter. Send:
- Links to your stuff — portfolio, GitHub, homelab writeup, demos. Whatever best shows what you build when nobody's asking.
- A short writeup (one page max) on one project where AI did most of the typing: how you and the model built it, and how you caught it.
- Your resume, one page.
US
