Research Scientist / Engineer, Alignment & Reward Modeling (all genders)
Core
Building reward models, LLM judges, and evaluation frameworks to ensure AI agents are reliable, honest, and aligned with instructions in regulated public-sector work.
Role type
Senior Research Scientist / Engineer (Alignment & Reward Modeling)
Builds
Reward models, LLM judges, and safety/reliability evaluation pipelines integrated directly into RL training loops.
Domain
AI Safety, Alignment, Reinforcement Learning, Public Sector Technology
Deliverable
production ML models
Required skills
Reward modeling, RLHF, preference learning, LLM-as-judge, alignment, safety evaluation, Python, PyTorch, metric skepticism, experimental design
Preferred skills
Interpretability, scalable oversight, publications on reward modeling, regulated industry experience, EU AI Act familiarity
Technologies
PyTorch, Python, EuroHPC, commercial compute
Responsibilities
Build rubric-based generative reward models and LLM judges; hunt reward hacking and specification gaming; conduct alignment experiments in RL post-training; perform safety and reliability evaluations with real-world cases.
Seniority
Senior, hands-on IC
