Thinking Machines Lab
Research Lead, Tinker, Fine-tuning Science
$350K–$475K
San Francisco · On-site · Full-time · Posted Sep 16, 2026
- Education
- Bachelor’s degree or equivalent experience in a related discipline; PhD or equivalent industry research experience preferred.
- Visa sponsorship
- Visa sponsorship is offered; success is not guaranteed for every candidate or role.
About the job
Leads the Fine-tuning Science team, setting research direction for frontier model customization and Tinker while advancing fine-tuning and post-training research and shipping findings into the product.
What you'll do
- Set the research agenda and roadmap for Tinker’s quality, efficiency, and reliability.
- Hire, mentor, and develop researchers while remaining hands-on in research.
- Advance fine-tuning and post-training techniques, improve large-scale run stability, and share findings through papers and technical writing.
What they ask for
- Bachelor’s degree or equivalent experience in Computer Science, Machine Learning, Physics, Mathematics, or a related discipline, with strong theoretical and empirical grounding.
- Track record of leading research and delivering on a team or major research effort.
- Proficiency in Python and familiarity with deep learning frameworks; able to debug distributed training and write scalable code.
- Clear technical communication and ability to build alignment across science, systems/infra, and product.
Nice to have
- Experience with RLHF, RLAIF, preference modeling, or reward learning for large models; or RL training stability techniques.
- PhD in a related discipline or equivalent industry research experience; contributions in alignment, data-centric AI, or human-AI collaboration are also valued.
Pay
$350,000 - $475,000 USD (annual salary)
Tools
Python, PyTorch, TensorFlow, JAX, LoRA
AI focus
Research on fine-tuning, frontier post-training, parameter-efficient fine-tuning, and RL to enable custom models; improve Tinker’s large-scale training quality, efficiency, stability, and reliability.
Model-written summary · the original posting is authoritative