Description
Thomson Reuters Labs is seeking a Lead Research Scientist to drive research on agentic reinforcement learning (RL). You will work on finding new, efficient ways to obtain reliable training signals for complex, multi-step tasks in professional domains such as legal, tax and compliance.
You will work with product and engineering teams across Thomson Reuters, with academic partners at world-leading research institutions, and with subject matter experts who have decades of experience in data collection, testing and evaluation of trained models.
Key Responsibilities
- Lead research direction: Own a research agenda in agentic RL and reward design, identify which emerging methods are worth pursuing, and turn them into a roadmap for our models.
- Innovate: Invent dense, reliable and cost-efficient training signals for real-world tasks where ground truth is scarce, delayed or partly subjective.
- Experiment rigorously: Design ablations and diagnostics that separate algorithmic effects from data, reward and system artefacts, then carry results into production models.
- Publish: Publish regularly as lead or senior author at top-tier venues and represent Thomson Reuters Labs in the research community.
- Mentor and collaborate: Guide scientists and engineers, and partner with Research Engineers to turn research ideas into trained models at scale.
- Communicate: Share findings through seminars, lectures, conferences, publications and released technical assets.
Required Qualifications
- PhD in machine learning, computer science or a related discipline.
- 3+ years of research experience after the PhD, in industry or academia.
- Strong publication record at top-tier venues on RL for LLMs, reward modelling, reasoning, agents and tool use, or preference optimisation.
- Hands-on experience post-training LLMs with RL or preference optimisation, using PyTorch (Advanced).
- Proven experience mentoring or supervising researchers, engineers or students, and guiding their work to published or deployed results.
Benefits
- Hybrid Work Model: Flexible hybrid working environment.
- Flexibility & Work-Life Balance: Supportive workplace policies, including work from anywhere for up to 8 weeks per year.
- Career Development and Growth: Culture of continuous learning and skill development.
- Industry Competitive Benefits: Comprehensive benefit plans, including flexible vacation, Mental Health Days off, access to Headspace app, retirement savings, tuition reimbursement, and employee incentive programs.
For more fresh new jobs, check out yubhub.co.