AI engineer (AI & GenAI) - People Domain (f/m/d)

As part of the opening of a permanent position, we are recruiting an AI engineer (f/m/d), based in Paris.

🚀 CONTEXT & STRATEGIC CHALLENGES

Within Decathlon Digital, the People Data team is launching a major initiative: to transform the experience of our 100,000 employees by integrating Generative and Agentic AI.

Faced with the fragmentation of the HR ecosystem in over 70 countries, our ambition is to build a simplification layer by leveraging the new solutions offered by AI. This will be a centralized and tool-agnostic intelligence layer, capable of interacting with our entire infrastructure (Workday, Data Lake, University, ...) to evolve our interfaces from a consultation logic to an action logic.

Within this framework, we are developing DaiLY, our HR conversational agent. It helps our employees and managers to instantly access reliable information, sourced from our Workday documentation and People policies.

Already successfully deployed in France, DaiLY is entering a new scaling phase: international expansion, industrialization of evaluations, continuous improvement of our models and their quality, and deployment and integration of autonomous HR agents. 

To lead this project, we are not just looking for a developer; we need a 'Builder': someone capable of translating business needs into technical solutions, implementing them, qualifying them, deploying them, and maintaining them over time.

 🎯 YOUR MISSION

We are looking for an experienced AI Engineer to actively participate in the industrialization and development of DaiLY. Your objective is clear: to transform HR business needs into reliable, measurable, and robust AI functionalities.

To take on this challenge, you will work within a dedicated team, in direct collaboration with other AI Engineers, a Lead AI Engineer (who will guide the overall architecture choices), and a Product Manager. This close-knit organization will allow for perfect synergy between product challenges, the roadmap, and technical implementation.

You will play a key role in the evolution of DaiLY: moving from a simple consultation model (Q&A) to a true transactional agentic system capable of triggering concrete actions (e.g., requesting leave in Workday, enrolling in training via Decathlon University).

🛠 YOUR RESPONSIBILITIES 

1. Agentic Development & Functionalities

• Participate in extending DaiLY's capabilities to write actions by implementing tool-calling protocols (MCP - Model Context Protocol) and inter-agent communications (A2A).
• Implement advanced chunking and re-ranking strategies to prioritize HR-specific rules.
• Contribute to optimizing costs, latency, and overall system efficiency in conjunction with the Platform and Infrastructure teams.

1. Quality, Evaluation & Observability (LLMOps)

• Develop and maintain the quality layer of the multi-agent RAG pipeline (Vertex AI Search / Discovery Engine).
• Ensure the grounding of responses and verification of citations to guarantee traceability to a source of truth.
• Implement robust observability pipelines for online and offline evaluation (OpenTelemetry, MLflow).
• Code automatic evaluators based on golden datasets to iterate with confidence and prevent regressions before each production release.
• Integrate guardrails to protect the system and users against hallucinations and errors.

1. Security, Isolation & International Deployment

• Access Management (ACLs): Implement granular and dynamic access rights management at the indexing and retrieval level, ensuring that an employee strictly accesses only the HR documents within their scope.
• Isolation & Multi-tenancy: Implement metadata filtering strategies in the vector database and isolation (of indexes and/or models) to ensure perfect data separation between Decathlon's different fiscal companies, legal entities, and geographies.
• Multilingual Support: Contribute to the internationalization architecture of DaiLY by adapting RAG pipelines, embedding, and inference to maintain a high level of performance and relevance across a wide diversity of languages.

1. Product & Business Impact

• Assist in analyzing user intentions to measure DaiLY's impact.
• Translate business challenges into model behaviors and success metrics, with the support of the Product Manager and Lead.

🧠 PROFILE REQUIRED

We are looking for a candidate with at least one initial experience.

• Experience: Experienced profile with several years of professional experience in a company (excluding internships and work-study programs) in AI Engineering, ML Engineering, or AI-oriented Software Engineering. You also have at least one concrete and successful experience in deploying a GenAI or agentic system into production.
• Technical Stack: Solid command of Python, the GCP / Vertex AI ecosystem, and orchestration frameworks (LangChain, LlamaIndex, or ADK).
• LLM & Agentic Expertise: Solid command of RAG architectures (vector search, Vertex AI Vector Search, advanced metadata management), but above all, agentic technical concepts: function/tool-calling, reasoning loops (ReAct, etc.), multi-agent workflow orchestration, structured output, and context/memory management.
• Quality & Evaluation: Ability to diagnose an incorrect response (model, retrieval, or data issue) and an aptitude for application observability (SQL / BigQuery / MLflow).
• Languages: Fluent French required & fluent professional English (international environment).

The pluses:

• In-depth technical knowledge of A2A (Agent-to-Agent) and MCP protocols.
• Experience in multimodal document processing or in HRIS ecosystems (e.g., Workday).
• Technical understanding of security challenges (RBAC, ACLs, GDPR) applied to LLMs.
• Knowledge of microservices architectures and API design.
  
  

🎁 WHAT WE OFFER

• The opportunity to develop a cutting-edge agentic AI product with a direct impact on the daily lives of 100,000 employees.
• A premium technological environment (Google Cloud / Vertex AI / Databricks Stack).
• A flexible remote work policy (2 days per week remote, the rest in our Paris offices - Canal de l’Ourcq).
• A delivery culture focused on autonomy, collective intelligence, and skill development (support from a Lead, certifications, training).
• Attractive benefits (monthly/quarterly bonuses, employee share ownership).
• An atmosphere resolutely focused on sport (numerous company sports events, product testing, sports or non-sports lunch breaks, …)