Europe or Latin America · Remote · Full-time
AI Software Engineer
Rose is our AI agent for OpenTelemetry® instrumentation. It reviews customer code for instrumentation problems, opens pull requests that fix them, adds baseline instrumentation where there is none, and migrates services off proprietary agents. We also publish open source agent skills, used by Rose and by coding agents everywhere, that give agents accurate, current OpenTelemetry knowledge. You will build Rose with a small team in a TypeScript codebase spanning a Cloudflare Workers control plane, sandboxed agent runs, and our own agent harness, and you will own those skills.
What you’ll own
- Build Rose features from spec to production, starting with GitLab support and a flow that turns findings from our telemetry analysis into fixes across a customer’s repositories
- Make agent runs reliable: structured output, retries, long-running sandbox jobs, and stable findings from one run to the next
- Build the evaluations that tell us whether a prompt or model change made Rose better, and what it cost
- Engineer context for agents that review and change Go, Java, Kotlin, TypeScript, Python, and .NET code
- Maintain our open source agent skills: keep them current with each OpenTelemetry release, review community contributions, and require evals showing that each change helps agents
- Work with customer engineers when a delivery needs it, together with our Customer Success team
Outcomes that matter
- Customers who use GitLab can run Rose on their repositories
- Every prompt or model change to Rose comes with evaluation results for quality and cost
- Our open source skills keep pace with OpenTelemetry releases, and community contributions get a timely, well-reasoned review
- Rose delivery work runs without pulling the rest of the team off product work
What we’re looking for
- Senior engineering experience, including an LLM agent or AI-assisted product you shipped to production and kept running
- Strong TypeScript and backend skills: APIs, webhooks, durable workflows, and Postgres
- Evidence that you measured LLM quality with evals, golden sets, or production signals, and changed course based on the results
- Comfort reading and safely changing code in languages you don’t write every day
- Clear writing and an async-first way of working: design docs, PR descriptions, and summaries
- Based in Europe or Latin America
Bonus points
- Experience instrumenting services with OpenTelemetry, or working on observability tooling
- Experience maintaining an open source project, especially in the OpenTelemetry ecosystem
- Experience writing agent skills, plugins, or MCP servers for coding agents
- GitHub App or GitLab integrations, sandboxed code execution, Cloudflare Workers, or Bun
- Enthusiasm for sharing your work in public, from blog posts to conference talks
Interview process
Show us what you built
The process is practical. We want to see agent systems you have shipped, how you knew they worked, and how you think about a real problem from our codebase.
- 01
A conversation with an engineer from the Rose team about your background and why OllyGarden interests you
- 02
A technical deep dive into an agent system you shipped: architecture, failure modes, and how you measured quality
- 03
A paid, self-contained exercise based on a real Rose problem, such as adding an evaluation or sketching a source-control integration
- 04
A conversation with our CEO, Juraci Paixão Kröhling, followed by reference calls