Python AI Engineer (Claude Code · RAG · Voice)

What you'll do:

You will design, build and ship LLM-powered features end-to-end. Day-to-day this means owning a problem from a Linear ticket through to a deployed Python service in production: prompts, evals, retrieval pipelines, agentic workflows, integrations with our CRM data, and the operational tail of monitoring and on-call.

You will work directly with the founder. There is no PM filter and no architect filter. You decide on the design, you write the code, and you run the deploy. AI tools (Claude Code, Cursor, MCP servers) are part of the workflow — we expect you to leverage them aggressively, not babysit them.

Requirements:

- 3+ years of production Python (FastAPI / async / typed code)

- 30+ days of daily Claude Code usage with a public /insights HTML report you can share

- Demonstrated RAG / agentic systems experience (vector DBs, embeddings, tool-use, function-calling)

- Comfort with PostgreSQL, Docker, and at least one cloud (AWS / GCP / Hetzner)

- Strong written English (B2+) — async-first team, written communication is the default

- Self-managed: we do not assign you tasks daily, you pull from a backlog

Strong plus:

- LangChain / LlamaIndex / DSPy production experience (not tutorial level)

- MCP server authoring or skill/hook engineering on Claude Code

- LLM evaluation infrastructure (custom evals, regression suites)

- Open-source contributions in the LLM ecosystem

- Experience scaling RAG above 10M documents

Selection process:

1. Application form with your /insights HTML report (filtering happens here)

2. 48-hour async test — small RAG service from a starter repo we send you

3. 45-min interview — we discuss the test, your /insights, and the trade-offs you made

4. Offer — monthly retainer USD 1500-6000 depending on level and timezone overlap

We do not run live coding interviews and we do not run "culture-fit" calls.

A note on language

Day-to-day internal communication (chats, calls, design discussions) happens in Ukrainian and Russian. Code, PRs, tickets, documentation, and any client-facing work stay in English. We are upfront about this so candidates can opt in or out before applying — being comfortable working in a Ukrainian/Russian-speaking team is a hard requirement for the role.