Senior AI Data Scientist — Agentic Process Development

As Senior AI Data Scientist you will build agentic systems to run HR processes end to end across multiple countries. You’ll map processes, quantify costs, and design stateful, production-ready agents that can trigger actions and integrate with various systems. The role emphasizes robust, deterministic automation over pure modelling, with safety gates, rollbacks, and governance. You will own end-to-end delivery—from concept to containerisation, instrumentation, and maintenance. This is a high-impact, cross-functional role at scale, shaping how HR operations are automated across Europe.

• Map leaver offboarding across 19 countries and translate into a process inventory with cost per step
• Facilitate workshops with HR, Rewards, and Ops leads to assess automation candidates for impact and feasibility
• Design state transitions, triggers, branches, system calls, wait points, and failure handling for multi-step flows
• Implement guardrails: dry-run mode, approval gates, least-privilege credentials, rollback paths
• Determine human-in-the-loop points, confidence thresholds, and a practical review queue
• Develop evals for output beyond accuracy (task completion, hallucinations, bias testing)
• Wire agents to webhooks with idempotent handlers to avoid duplicate actions
• Decide which steps require frontier models vs. cheaper alternatives and prove routing decisions with data
• Evaluate vendor APIs, data models, integration costs, and API exposure in vendor demos

• 7+ years building data and ML systems in industry
• Expert in Python and ML; production-ready code and containerisation
• Experience shipping end-to-end multi-step LLM workflows with orchestration and fault tolerance
• Cost and latency engineering mindset: model routing, caching, batching
• Strong governance and safety practices: least-privilege, approval gates, rollback
• Ability to map real processes with quantification and stakeholder facilitation
• Executive-level business case writing and cost modelling

• Facilitation with senior stakeholders
• Stakeholder management and clear communication
• Pragmatic decision-making under high risk
• Python with solid readability
• Docker or equivalent containerisation
• End-to-end ML systems with multi-step tool-calling workflows