Ph.D. / Postdoctoral / Master’s Expert

About this opportunity

Turing is seeking Master’s, Ph.D., and postdoctoral-level STEM experts to work on projects that help fine-tune and evaluate large language models. The role involves designing challenging STEM problems, producing clear step-by-step solutions, collaborating with LLM researchers, and helping define evaluation benchmarks across topics ranging from early undergraduate to Ph.D.-level material.

What the listing describes

• ✓Design challenging STEM problems to probe the limitations of large language models
• ✓Solve complex STEM problems and create clear, high-quality step-by-step solutions
• ✓Provide well-articulated reasoning and detailed annotations
• ✓Collaborate with LLM researchers to align problems with evaluation goals
• ✓Focus on areas where language models typically struggle, including abstraction, multi-step reasoning, and symbolic manipulation
• ✓Help define new evaluation benchmarks based on Physics curricula
• ✓Work with topics spanning early undergraduate through Ph.D.-level material
• ✓Provide constructive feedback on model outputs and evaluation tasks

Requirements shown on the source

• ✓Strong research and analytical skills
• ✓Ability to provide constructive feedback and detailed annotations
• ✓Creative and lateral thinking abilities
• ✓Excellent structured communication and collaboration skills in a remote setting
• ✓Self-motivated and able to work independently
• ✓Desktop or laptop with a good internet connection
• ✓Strong STEM foundation, particularly at the level expected in engineering entrance exams and graduate or Ph.D.-level programs
• ✓Ability to break down complex STEM concepts into simple and clear explanations

Relevant backgrounds mentioned

• ✓Currently pursuing a Master’s, Ph.D., or postdoctoral degree in STEM, Applied Physics, or a related field
• ✓Ability to analyze and solve complex Physics problems using a structured and logical approach
• ✓Ability to explain STEM concepts clearly using simple language, visuals, and Physics reasoning