Machine Learning: Multimodal Foundation Models

The role

You will build unified foundation models that reason across text, image, video, and kinematics to drive intelligent robotic policies. You will work on large multi-modal networks and own the entire stack from data to training and deploying models.

What you will do

• Build Native Multimodal Policies by developing architectures where vision, language, and more modalities share a unified representation.
• Improve Cross-Modal Reasoning by researching and implementing methods to ensure the model reasons through modalities.
• Own the Training Loop End-to-End by designing, running, debugging, and iterating on large-scale training experiments.
• Ship and Iterate on Real Systems by integrating models into real robotic stacks and optimizing performance for edge inference.

Required qualifications

• Very strong coding skills in Python, C++, or Rust.
• Production MLLM Experience: Track record of training and deploying large-scale multimodal models.
• Pretraining & RL Mastery: Deep intuition for LLM-style pretraining, post-training, and Reinforcement Learning at scale.
• Infrastructure Fluency: Comfortable managing and optimizing large-scale experiments on massive GPU clusters.
• Exceptional mental acuity: you think quickly, learn instantly, and reason across unfamiliar domains.
• Engineering curiosity: you naturally dig into how systems work, even outside your specialty.
• High performance mindset: you move fast, handle ambiguity, and excel when the environment is demanding.

About the company

The Bot Company is a small team of engineers, designers, and operators based in San Francisco building a helpful robot for every home.