Research Mid Training
Artificial-intelligence research and product company building customizable AI systems, including the Tinker training API and Inkling open-weight models.
Funding history
About Thinking Machines Lab
Thinking Machines Lab develops AI products that let researchers and developers fine-tune and use models, while also releasing open-weight multimodal models.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will own late-stage model training across data strategy, data quality, training recipes, and evaluation. You will source, curate, synthesize, filter, and verify training data; develop quality systems and model interventions; tune training configurations; measure outcomes; debug results; and explore scalable training methodologies.
Requirements
- Proficiency in Python
- Familiarity with PyTorch, TensorFlow, or JAX
- Ability to debug distributed training and write scalable code
- Bachelor’s degree or equivalent experience in Computer Science, Machine Learning, Physics, Mathematics, or a related discipline
- Ability to explain complex technical concepts in writing
Responsibilities
- Source, curate, synthesize, filter, deduplicate, verify, and rewrite training data
- Design and measure knowledge-improving interventions
- Introduce and measure model behaviors during mid-training
- Build automatic filtering, scoring, verification, and rewriting systems
- Develop and tune datasets, training stages, annealing schedules, and hyperparameters
- Define and iterate on evaluations
- Debug training configurations and develop deeper understanding
- Measure performance scaling and explore new training-data methodologies
Benefits
- Health benefits
- Dental benefits
- Vision benefits
- Unlimited PTO
- Paid parental leave
- Relocation support