AI Engineer (Agents)

About the role

Build how our agents think: understand a request, plan, use tools, ask for approval when needed, and not make things up when they do not know.

What you will do

• Design and build agents and multi-agent workflows (Mastra, Vercel AI SDK).
• Implement tool calling, structured outputs and MCP, with human approval for sensitive actions.
• Build evals that stop quality regressions, and run them on every change.
• Choose and route between models (OpenAI, Anthropic, Google) by quality, speed and cost, and measure it (Langfuse).

What matters to us

• 4+ years in software engineering; TypeScript required, Python a plus.
• At least one LLM product shipped to real users, with evals and failure handling.
• Deep understanding of tool calling, structured outputs, RAG and memory.
• Measurable thinking: "how will we know it works?".

Nice to have

• Experience with MCP.
• Postgres and pgvector.
• Monitoring AI systems in production.

What you will work with

• TypeScript
• Python
• Mastra
• Vercel AI SDK
• MCP
• OpenAI
• Anthropic
• Google
• Langfuse
• PostgreSQL
• pgvector
• Inngest

Your first 90 days

1. 30

   Month one: learn the agent engine, and build a first eval for a real scenario.

2. 60

   Month two: improve one scenario end to end, and measure the gain.

3. 90

   Month three: lead a new agent capability, from the design until it reaches customers.

About Troupia

We are building a team of AI agents that does the day-to-day work of businesses and organisations, beside the people in them. The product is live, in eight languages, and the team is small: whoever joins now shapes where it goes.

How we hire

1. 1A short intro call.
2. 2A professional conversation with the people you will work with, sometimes with a short task.
3. 3A meeting with the founder.
4. 4An offer. And at every step, you get an answer, even when it is no.