Agentic AI Engineer

Il y a 23 heures

Paris, Ile-de-France Culligan International Temps plein
About Culligan For 90 years, Culligan has helped people and businesses enjoy better water. Present in 90 countries, we are a global leader in water treatment and services, offering water softeners, purifiers, dispensers and related services for homes and businesses. Our purpose is to transform water to improve everyday life and help protect the planet. Our mission is to provide quality water through three strengths: deep water expertise, premium service and environmentally responsible solutions. We are a service business at heart, with an ambition to earn customers' trust through outstanding service and highly engaged teams. About the EMEA AI Foundry We're building a Paris-based team to rethink how Culligan works across EMEA. Customer experience is one of our starting points; the ambition extends to commercial and operational transformation. The Foundry brings together product, engineering, data and business expertise to design, ship and scale AI solutions alongside country teams and IT. Our first deployments focus on customer experience in Germany, Spain and Portugal. We are also building shared agentic, data and knowledge foundations so that what works in one country can be adapted and scaled across others. The challenge is tangible: complex service operations, multiple systems and real people who depend on the solutions we deliver. How we work: Set the standard. Ship what matters. Practical genius. Win together. Why this role You will build AI agents that move beyond demos into real customer and operational workflows. The engineering challenge is substantial: fragmented enterprise systems, imperfect data, tool use, evaluation, security and reliable production operation. You'll help build the framework while shipping agents on it, with a direct view of how they perform in the business. The role Build and operate agents end to end: their behaviour, tools, context, data and the evidence that they work. We are hiring two complementary profiles
- one stronger in applied AI and data science, the other in software and data engineering. Both have the same production ownership, contribute across the stack and learn from each other. What you will own
• Agent design and delivery
- implement prompts, context, orchestration, tool use, guardrails and safe hand-offs using the shared framework.
• Evaluation
- build representative test sets, score behaviours, analyse failures and make evaluation a release gate, including for agents that act on business systems.
• Integrations
- build secure, reusable APIs, connectors and tool contracts for enterprise systems, using MCP where it adds value and other patterns when appropriate.
• Data and knowledge
- improve pipelines, retrieval, grounding and quality with the Data Product Architect.
• Applied AI and optimisation
- contribute models for intent, routing, churn, next-best action or operational planning when the use case calls for them.
• Production operation
- instrument agents for reliability, latency, quality and cost; support incidents, safe rollback and continuous improvement alongside IT operations.
• Business partnership
- work directly with Product Managers and country teams to understand the workflow, test with users and make pragmatic design trade-offs. Between the two roles, we expect complementary depth: applied AI, evaluation and modelling on one side; integrations, data pipelines, delivery and operations on the other. The stack We use current tools and expect engineers to help improve the choices as the platform matures:
• Azure Databricks and Unity Catalog for data, pipelines and governance.
• Microsoft Foundry and LangGraph for agent development and orchestration.
• MCP and other secure integration patterns for tools and enterprise APIs.
• Frontier LLMs, starting with OpenAI models.
• Python, Azure DevOps for source control and CI/CD, and Bicep for infrastructure as code. Where you have the final call You own implementation decisions and the technical evidence of release readiness within standards set by the Agentic Platform Architect. You can block a release that fails the agreed quality, safety or reliability gates. Product Managers own functional acceptance; production go-live is agreed with country business owners and IT. Your first 90 days
• Take an initial agent end to end into a controlled production release, working with a Product Manager and the country teams.
• Build and use an evaluation suite as a real release gate, including failure analysis and safe fallback behaviour.
• Contribute a reusable integration or tool interface and establish monitoring, cost visibility and the operational handover. What we are looking for
• Senior hands-on

experience:
you have shipped software or ML systems to production, operated them and can work with autonomy.
• Depth in either applied AI and data science, or software and data engineering; curiosity and working fluency across the boundary.
• Practical LLM

experience:
agent workflows, tool calling, retrieval, evaluation and debugging real failures.
• Strong Python skills, cloud fluency and an understanding of production quality, security and observability.
• For applied AI profiles: sound statistics, experiment design and model evaluation. For engineering profiles: strong API integration, service or pipeline operations and delivery practices.
• Enough product sense to question requirements and explain technical choices in terms of user and business impact. Azure or Databricks experience is a plus. How success is measured Specific objectives and baselines are agreed during your first quarter. We value production impact and engineering quality over the number of demos or agents shipped.
• Value in production
- agents or shared capabilities reach real users and contribute to agreed business outcomes.
• Quality and safety
- evaluation coverage, resolution quality, safe actions and appropriate human escalation improve over time.
• Reliability and efficiency
- production incidents, recovery, latency and cost per successful interaction are monitored and improved.
• Reusability
- sound technical components and integrations are adopted across use cases when reuse genuinely makes sense. Practical details Location Paris, with regular travel to the EMEA countries where we deploy. Contract Permanent, full-time. Language English is the working language of the team. Reporting Agentic Platform Architect. Partners Agent Product Managers, Data Product Architect, country businesses and IT, which owns infrastructure and security policy. Compensation Discussed individually, based on profile and experience.

How to apply
Send your CV to sami.taaissat@culligan.com. We use four steps and aim to give you a clear answer after each one. 1. Agentic Platform Architect
- meet your future manager and discuss production experience and day-to-day engineering work. 2. Head of Agentic Transformation
- discuss business value, delivery priorities and working with countries. 3. Technical case with the Agentic Platform Architect
- work through an agent design, evaluation and production-readiness problem. 4. VP AI & Digital Transformation EMEA
- discuss the broader ambition, fit with the team and final decision.
• Culligan is an equal opportunity employer. We assess applications on skills and evidence, and welcome candidates from every background.