Research, Mid Training
$350k - $475k • San Francisco • FullTime
Posted 26d ago
Job Location
San Francisco
Tech Stack
Remote Work Policy
On-site
Employment Type
FullTime
Categories
AI Research Engineer
About the job
Thinking Machines is building AI that extends human will and judgment, focusing on training frontier models, developing tools for model customization, and crafting interfaces for human-AI communication. This role is crucial for shaping the fundamental reasoning capabilities of our models by owning the late-stage training process. It involves a blend of fundamental research and practical engineering, making it ideal for individuals comfortable working at the intersection of pre-training and post-training.
Responsibilities
- Own the data strategy, including sourcing, curating, synthesizing, and building pipelines for training-grade datasets that reflect real-world model usage.
- Design and measure interventions to improve model knowledge, such as targeted corpora, synthetic rephrasing, and knowledge-dense mixes.
- Introduce and measure new behaviors during mid-training, assessing their impact on post-training efficiency and stability.
- Build and own the automatic filtering and scoring stack for data quality, including training classifiers and LLM judges.
- Develop and tune mid-training recipes, including datasets, training stages, annealing schedules, and hyperparameters, and measure their effects.
- Iterate on evaluation methodologies to ensure they accurately capture meaningful model performance improvements.
- Debug and understand unexpected results in training configurations to improve model performance and deepen understanding.
- Scale existing methodologies and explore new ones for mid-training, measuring performance scaling with dataset size and exploring different training data types.
Requirements
- Proficiency in Python and familiarity with deep learning frameworks like PyTorch, TensorFlow, or JAX.
- Comfort debugging distributed training and writing scalable code.
- Bachelor’s degree or equivalent experience in Computer Science, Machine Learning, Physics, Mathematics, or a related discipline.
- Clarity in communication and ability to explain complex technical concepts in writing.
- Strong grasp of probability, statistics, and ML fundamentals.
- Experience building or owning training datasets for large models.
- Experience with knowledge injection, continued pre-training, or domain adaptation of large models.
- Experience building model-based data quality systems.
- Experience generating synthetic data and understanding its failure modes.
- Research or engineering contributions in alignment, data-centric AI, or human-AI collaboration.
Benefits
- Generous health, dental, and vision benefits
- Unlimited PTO
- Paid parental leave
- Relocation support