Thomson Reuters
Thomson Reuters
Thomson Reuters develops trusted content, data, and technology for professionals working in legal services, tax and accounting, international trade, risk management, insurance, consulting, and global news. Its portfolio includes legal research platforms, tax and accounting software, compliance and regulatory tools, and Reuters media services. By combining specialist expertise with investments in generative AI, Thomson Reuters helps law firms, corporations, tax professionals, and media organizations manage complex information, streamline workflows, and make better-informed decisions.

Director, Model Research and Development — Zug, Switzerland (Hybrid)

Lead Thomson Reuters’ LLM post-training, reinforcement learning, and data strategy. Develop reliable AI models for legal, tax, and other regulated professional workflows.

Description

  • Provide technical leadership for the model research program
  • Advance post-training and data strategies for accurate, calibrated, and trustworthy models used in professional workflows
  • Direct hands-on post-training of Thomson Reuters’ LLMs through supervised fine-tuning, preference optimization, and reinforcement learning
  • Build and operate online, agentic reinforcement-learning pipelines with subject-matter experts involved
  • Direct data selection, mixture optimization, synthetic-data creation, and evaluation design
  • Track training runs, detect issues early, and define appropriate corrective measures
  • Partner daily with infrastructure and evaluation teams to sustain dependable training and assessment pipelines
  • Present research findings and recommendations to guide roadmap and prioritization decisions
  • Contribute to experiments, training and evaluation pipelines, and research papers for leading conferences
  • Report to the Vice President of Model Research & Development

Requirements

  • Demonstrated expertise in supervised fine-tuning, preference optimization such as DPO, and reinforcement learning, including agentic multi-step reinforcement learning for tool use and task completion
  • Practical experience defining data selection, mixture design, synthetic-data generation, and evaluation approaches, with measured impact on model behavior
  • Strong hands-on expertise in distributed training, data pipelines, and evaluation infrastructure, including the ability to build and troubleshoot these systems
  • Record of shipped models, broadly adopted open-source work, or peer-reviewed publications at leading venues such as NeurIPS, ICML, ACL, or EMNLP
  • Experience leading a focused technical team responsible for training and/or evaluation
  • Experience applying LLMs in law, tax, or a similarly regulated professional field preferred
  • Experience with continued pre-training and domain adaptation for large base models preferred
  • Experience in model safety, adversarial robustness, and red-teaming preferred

Benefits

  • Flexible hybrid work arrangements for office-based roles
  • Up to eight weeks of working from anywhere each year
  • Flexible vacation policy
  • Two company-wide Mental Health Days annually
  • Access to the Headspace app
  • Retirement savings program
  • Tuition reimbursement
  • Employee incentive programs
  • Resources supporting mental, physical, and financial wellbeing
  • Two paid volunteer days each year
  • Opportunities to join pro bono consulting projects and ESG initiatives
  • Career development, ongoing learning, and skills training through Grow My Way
  • Flexible workplace policies designed to support work-life balance
  • Employee inclusion, belonging, and social impact programs

Related Jobs

Thompsons Solicitors

Clinical Negligence Lawyer

Thompsons Solicitors
501 – 1,000 Employees
B2CLegal

Handle complex clinical negligence claims at Thompsons Solicitors, a UK social justice law firm, and guide junior lawyers. Help injured clients pursue justice through high-quality litigation and client care.

Open
T

North Carolina Foreclosure Attorney

Timothy Padgett
11 – 50 Employees

Handle foreclosure matters for Padgett Law Group by drafting legal documents and coordinating case logistics. The role supports institutional financial clients and includes preparation for mediation, court, and Sheriff’s Sales.

Open
T

Remote Louisiana Bankruptcy Attorney

Timothy Padgett
11 – 50 Employees

Represent institutional creditors in Louisiana bankruptcy proceedings, handling hearings, pleadings, case dockets, and legal guidance. Support Padgett Law Group with case-law compliance and timely communication with clients and staff.

Open