- San Francisco, California, United States
- On-site
- Full-time · Mid
- 3+ years
We’re looking for a motivated individual to design, implement, and refine efficient Large Language Models (LLMs) pipelines for scaled distributed training, focusing on reasoning algorithms that go beyond traditional LLMs.
Responsibilities
- Implement new reasoning algorithms and models
- Evaluate reasoning approaches, including latent space reasoning
- Pre-train, fine-tune, and modify the State-of-the-Art LLMs
- Optimizing and scaling LLM pipelines
- Adjust frameworks and interfaces to accelerate machine learning development
Requirements
- Nice to have: torch.compile
- Nice to have: EBM
- Nice to have: Latent reasoning
- Nice to have: Mathematical Reasoning
- Nice to have: Discrete math