L'IA agentique d'entreprise.
À la française.
TL;DR: You deploy the Idun Platform at large enterprise clients and co-build their first production agents side-by-side with their engineering teams. This is hands-on work — not advisory, not training. You configure the platform, write the first agents, define the delivery patterns, and leave the team genuinely autonomous. Paris-based, hybrid, €60–85k + equity.
Idun Group builds the operating foundation for enterprise AI agents. We develop an open-source runtime — Idun Engine — and a proprietary control plane — Idun Platform — designed to put agents in production on client infrastructure with centralised configuration, observability, tracing, memory, guardrails, access control, and multi-framework interoperability. Our clients include major European enterprises across aerospace, luxury goods, and industrials.
Most enterprise AI projects stall not at the model layer but at the infrastructure layer. The data team has a prototype; the platform team has no idea how to operate it; the security team has a list of blockers that nobody has worked through. We exist to close that gap.
As an AI Platform Engineer, you join an engagement as the technical anchor. You deploy the Idun Platform into the client's environment — cloud or on-prem, whatever their constraints demand — and you build the first agents with their team, not for them. The difference matters. By the end of an engagement, the client's engineers own the patterns, the deployment pipeline, and the production readiness checklist. You are the person who made that transfer real.
This is a senior-to-staff level role. You will hold technical conversations with Tech Leads, platform engineers, and infrastructure architects. You will push back when a proposed architecture doesn't scale. You will translate a vague "we want AI" mandate into a first working agent with proper observability, guardrails, and access control. The work is demanding and genuinely interesting — EU enterprises deploying agents under GDPR and the EU AI Act are solving problems that most of the industry has not figured out yet.
- Deploy and configure the Idun Platform in client environments, working around real security, network, and infrastructure constraints — cloud (GCP, Azure, AWS), on-prem Docker, air-gapped. - Co-build the first agents and first production use cases with client engineering teams using LangGraph, Google ADK, or Haystack. - Set up the full industrialisation stack: OpenTelemetry tracing to Langfuse or Phoenix, memory backends, guardrails policies, OIDC/RBAC integration, CI/CD pipelines, automated tests. - Lead technical scoping sessions — translate business requirements into agent architecture, data flow diagrams, and a deliverable component list. - Define engineering best practices for the client team: code organisation, configuration management, debugging playbooks, testing strategy, deployment runbooks. - Build reusable frameworks and templates that leave the client capable of extending the platform independently after the engagement closes. - Document deployment patterns, write post-engagement handoff guides, and contribute findings back into Idun's shared implementation knowledge base.
You will work across this stack on every engagement:
- Python, FastAPI, Pydantic, PostgreSQL — the core runtime and data layer - Docker, Git, CI/CD, REST APIs, YAML — configuration-driven deployment on any infrastructure - LangGraph, Google ADK, Haystack — the three primary agent frameworks we deploy against - MCP, A2A, streaming / SSE — protocol-level interop between agents, tools, and frontends - OpenTelemetry, Langfuse, Phoenix — the observability chain; Langfuse and Phoenix are the primary trace UIs - OIDC, RBAC, SSO — auth integration with whatever identity provider the client already runs (Google, Microsoft, ADFS)