TL;DR: Business teams at our clients prototype agents on Copilot Studio, Microsoft Foundry or Palantir AIP. You take them over and turn them into production services: architecture, secure integration, tests, deployment, documentation. Senior profile, at least 6 years of experience, at ease with business experts. Paris (Île-de-France), 50% on-site and 50% remote, €55–70k, permanent contract or freelance.
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 under a year across more than 10 large enterprise clients; our references include LVMH, Airbus and L'Oréal. We resell no platform, which leaves us free to pick whichever one holds up for each use case.
In large enterprises, agents are no longer born only in IT. A claims handler builds one in Copilot Studio, a finance team assembles another in Microsoft Foundry, a risk department tries Palantir AIP. The prototypes work in a demo. They do not reach production: no tests, loose access rights, no monitoring, nobody to maintain them.
Your job is to get them there. You start from a prototype and a business expert. You understand the process in question (underwriting, claims, pricing, finance, operations) and you deliver an agent the company can run: connected to its systems through secure APIs, tested, monitored, documented, with a team able to keep it alive.
You sit between tech, data, AI and the business. Half the role is engineering. The other half is understanding what the expert actually wants, turning it into behaviour the agent can hold, and checking that it does.
- Take over business prototypes and say which can go to production, under what conditions and on what timeline. - Design the target architecture: orchestration, tool calling, agent state and memory, RAG, skills and topics. - Connect agents to enterprise systems through REST APIs and secure integration patterns, and through MCP where it fits. - Write the tests and evaluation sets, set up monitoring, deploy. - Document and hand over: runbooks, development standards, coaching for less experienced developers. - Contribute to the client's shared standards for industrialising agents, whatever the platform.
- Microsoft: Copilot Studio, Microsoft Foundry, Azure - Other platforms: Palantir Foundry and AIP, Databricks - Language and integration: Python, REST APIs, enterprise authentication and access control - Agents: MCP, tool calling, orchestration, RAG, state and memory, prompting - Industrialisation: tests, evaluation, CI/CD, observability
- At least 6 years in AI, software engineering, data engineering or solution delivery on modern data and AI platforms. - You have shipped and operated data or AI applications in production, in an agile setting. - You have built generative AI applications, agents or tool-calling workflows, and at least one of them is running today. - Strong Python, solid knowledge of REST APIs and secure enterprise integration. - Hands-on experience with MCP or an equivalent standard for connecting agents to tools. - Industrial experience on at least one of Microsoft Foundry, Copilot Studio, Palantir AIP or a comparable enterprise platform. - You work directly with business experts and turn their needs into reliable solutions. Experience in insurance, reinsurance, financial services or another regulated industry is a real plus. - Analytical, accountable, keen to coach others. - Professional French and English: these environments are often international. - Based in Île-de-France and available soon. The role is split 50% on-site, at the client or in our Paris office, and 50% remote. This is a requirement of the role.
