AI Engineer, Developer Ecosystem
What you'll actually do
- Build agents and tools in public: demo apps, reference implementations, MCP servers, Claude skills, LangGraph workflows. Ship things that are genuinely impressive.
- Own the developer experience: identify friction in our API and SDKs, write real feedback back to the eng team, and fix it yourself when you can.
- Design and run evals: benchmark tool-calling quality, measure agent reliability across integration surfaces, build sandboxed test harnesses that reflect production conditions. Publish what you learn.
- Run workshops, give talks, appear at events: technical sessions on agentic architectures, tool-calling patterns, context optimization, and integration design.
- Publish AI research adjacent to your work: MCP tool schema design, context window hygiene, eval frameworks for agentic systems, RLMF, auto-research loops, sandbox architecture for safe agent execution.
- Foster community: Discords, GitHub, demo days, office hours. Be the engineer developers trust to give them a real answer.
- Partner with product and engineering: turn new releases into working demos before they're announced. No slide decks without code.
What we're looking for Hard skills
- Ship production-grade agents
- Deep MCP / tool-calling fluency
- Built plugins, skills, extensions, or agents for real usage
- Designs evals and benchmarks for agentic systems
- Builds sandboxes for safe agent testing
- Understands context optimization
- Reads AI research papers and applies them
- TypeScript and/or Python at minimum
Soft signals
- GitHub history you're proud of
- Technical talks on record
- Community presence
- Builds to learn, not to demo
- Gives direct opinions, backed by data
- Doesn't wait to be unblocked
What we're not looking for
- Someone who needs to ask permission to write a blog post or be taught on how to open a PR
- Someone whose agent experience is only a weekend hackathon project
- A conference talk collector with nothing on GitHub
Topics you should have opinions on MCP
- A2A protocol
- tool-calling schemas
- context window optimization
- evals & benchmarking
- agent sandboxes
- LangGraph / DSPy
- RLMF / RLM harnesses
- auto-research loops
- code mode / long-horizon agents
- RAG vs. tool-use tradeoffs
- enterprise auth for agents
- multi-agent orchestration
- prompt caching strategies
- AI safety boundaries
- sandbox isolation patterns
- LLM leaderboard literacy
This is a real engineering role This isn't a "write blog posts and attend conferences" role dressed up as engineering. You'll be embedded with the product and engineering team. You'll ship code that ends up in our SDKs, our docs, and our sample repos. The AI agent ecosystem is moving fast enough that the line between DevRel and R&D is blurring. We want someone comfortable sitting in that blur - writing a technical post about eval design for tool-calling reliability because they spent two weeks deep in it, building a sandbox harness to reproduce a flaky agent behavior, not because someone briefed them on a slide. You'll have access to a platform that connects agents to any other system safely while optimising token usage, and a mandate to show the world what's possible when those connections actually work well. Apply tot his job Apply To this Job