AI Agent & Harness Engineer

You're genuinely cracked at AI agents. You build agentic systems that actually work — tool-calling, RAG, evals, the whole loop — and you wield AI coding harnesses (Claude Code, Cursor, etc.) to ship faster than seems reasonable. You'll build real agent features into our product (tool-calling LLMs, model serving, guardrails, eval harnesses) and the internal agent tooling that makes the whole team multiples faster, plus enough full-stack glue to get your work in front of users. We care about what you've built, not your résumé — early-career, self-taught, and unconventional backgrounds welcome.

Full time, long term, REMOTE, 3-4k monthly.

##### Must-haves

* Real things you've built with LLMs/agents — show us GitHub, demos, or side projects (this matters most)
* Strong Python, plus enough full-stack (some TypeScript/React) to ship a feature end-to-end
* Deep fluency with the modern agent stack: tool/function calling, RAG, prompt & context engineering, MCP, and evals
* Fluent with AI coding tools and opinionated about getting real work out of them
* Resourceful, fast, and self-directed
* Strong written English; remote self-management
* Some daily overlap with U.S. Pacific business hours

##### Nice-to-haves

* Self-hosted model serving (vLLM), GPUs, quantization (FP8)
* Building eval frameworks, guardrails, or agent observability
* Multi-agent orchestration or developer-tooling / harness work
* Django/DRF, PostgreSQL, or geospatial/scientific domains
* You ship side projects for fun and can't stop tinkering with agents