Thomson Reuters Labs is hiring Senior- and Lead-level research scientists to develop and evaluate the Thomson family of large language models for high-stakes professional workflows. The research team improves open-weight models through continual learning, including value alignment, data-centric mid-training on curated proprietary data, DPO, and reinforcement learning, rather than training from scratch or relying on narrow fine-tuning. The evaluation group develops expert-annotated legal and tax benchmarks, agentic systems, and human preference studies with subject-matter experts. The team regularly publishes its research. Candidates should indicate which one or two areas of emphasis best fit their experience: agentic reinforcement learning, model training, or model evaluation. The agentic RL area focuses on planning, long-context reasoning, tool use, long-horizon tasks in real agent harnesses, multi-turn RL, policy optimization, reward modeling, and training-signal design. Model training covers continued pre-training and post-training of open-weight models, DPO and RL, training data and reward design, and forgetting and replay. Model evaluation includes benchmarks and targeted evaluations for complex legal and professional workflows, LLM-as-a-judge and rubric design, human preference studies with legal experts, and using evaluation results to inform training objectives. At Lead level, the role owns a research agenda, turns emerging methods into a roadmap for the models, and mentors scientists and engineers. The posting is for proactive recruitment and may be used to fill current openings or future vacancies. It is scheduled to close on March 31, 2027.
Stand Out From the Crowd
Upload your resume and get instant feedback on how well it matches this job.
Job Type
Full-time
Career Level
Senior
Education Level
Ph.D. or professional degree