About Idun
Idun Group is an independent AI consultancy. We measure how AI is actually used inside our clients, from the logs their platforms already produce, then help them industrialise what works. More than 40 AI assessments in a year; our references include LVMH and L'Oréal. It is tooled consulting: our software, Idun Cockpit, measures real usage, then we say what to industrialise, consolidate or cut. Idun is backed by Wilco, IMT Starter, Polytechnique X-UP and La French Tech.
The role
Idun Engine v0.6 collapsed three services into a single wheel and deleted 151,820 lines of code. The surface is clean. The next 12 months are about making the runtime genuinely production-hardened, developer-friendly, and community-ready.
This is not a maintenance role. You are the technical owner of the Engine. You decide how the plugin interface evolves. You define the observability contract between the runtime and external trace backends. You architect the memory abstraction layer. You write the docs that let a developer in Amsterdam or Warsaw deploy a production-grade agent in an afternoon without reading source code.
The work has two surfaces. The first is internal: building the platform features that Idun's consulting practice deploys at clients (LVMH, Airbus, L'Oréal, all of which run on Engine). The second is public: maintaining the OSS repo, triaging issues, reviewing external PRs, writing release notes that explain not just what changed but why. Both surfaces matter equally.
You will work closely with Geoffrey (Chief Product Officer, co-founder) on architecture decisions. As a founding engineer, your API choices will be load-bearing. They will be the patterns that ten future engineers build on top of.
What you'll do
- Own the core runtime loop: the FastAPI service layer, the agent execution model, streaming via AG-UI, and the configuration system that drives deployment from a single YAML file.
- Design and implement the plugin and integration interface: the seam between the Engine's runtime and external frameworks (LangGraph, Google ADK, Haystack), tool servers (MCP), trace backends (Langfuse, Phoenix, LangSmith, GCP Trace), and memory backends (PostgreSQL, SQLite, Vertex AI).
- Lead the observability pipeline: asyncpg COPY writer, OpenTelemetry span ingestion, the admin dashboard KPIs, and the trace viewer, so operators get a real production view, not a demo.
- Drive the developer experience: CLI ergonomics (
idun setup,idun serve,idun deploy), error messages that tell you what to fix, documentation that does not require reading source code. - Write, maintain, and evolve the public documentation (MkDocs-based): quickstart, how-to guides, API reference, architecture explainer, migration guides between versions.
- Triage GitHub issues, review external PRs, and manage the OSS release cadence: versioned releases with a clear changelog and semantic guarantees.
- Contribute architectural context to consulting engagements when client deployments surface Engine limitations that feed back into the roadmap.
Stack & environment
- Python 3.12, FastAPI, asyncio, Pydantic: the entire Engine is Python; async-first throughout
- OpenTelemetry: the core observability contract; spans flow to Langfuse, Phoenix, LangSmith, or GCP Trace
- PostgreSQL, asyncpg, SQLite: the primary memory and trace store backends
- LangGraph, Google ADK: the two primary agent frameworks; the Engine wraps both without requiring either
- MCP (Model Context Protocol): per-agent tool server governance; the Linux Foundation standard since 2025
- Docker, GitHub Actions, pytest: the CI/CD and testing pipeline
- MkDocs: the documentation site
- Next.js (bundled): the admin + chat + trace viewer UI, served by FastAPI from inside the wheel (you do not own the frontend, but you do own the FastAPI surface it talks to)
Profile we're looking for
- 5+ years of production Python engineering: you have shipped services that real users depend on at real scale. You understand async, you write tests before bugs, you read other people's error reports with genuine curiosity.
- Experience with OSS maintenance: you have either maintained a library with external contributors, or you have a strong public record (GitHub, blog, conference talk) demonstrating that you know how to ship in public and communicate about trade-offs. We care far more about demonstrated OSS instincts than a specific maintainer title.
- Distributed systems and observability depth: you have spent time thinking about tracing, span semantics, distributed context propagation, and what makes a trace viewer actually useful to an on-call engineer.
- API design sensibility: you have opinions about what makes a library a pleasure to use, and you can articulate the trade-off between flexibility and simplicity. You do not default to "expose everything" as an API philosophy.
- Based in Île-de-France, or ready to move there before you start: the role is split 50% on-site in Paris and 50% remote.
Why join this role
- Founding-level ownership. You are the primary author of an open-source project used by enterprises across Europe. Your architecture decisions ship to production at Airbus and LVMH. Few engineering roles at this stage offer this level of real-world impact and technical autonomy.
- Equity that matters. 0.5–1.5% equity is a genuine stake. We are transparent about the cap table.
- 50% on-site, 50% remote. Half the week with the team in Paris, where the architecture decisions get made; the other half from home, for deep work.
- An interesting problem at the right moment. The agent runtime ecosystem is consolidating. The decisions being made now about open standards (MCP, AG-UI, OpenTelemetry for agents) will define the category for years. You will be in the room where those decisions are made, and your work will be part of the record.