Machine Learning Engineer

About the Role

An early-stage company is building infrastructure to teach machines to act in the physical world. They gather real-world experience and run on-device reinforcement learning to turn it into robot capability, with the aim of taking repetitive physical labour off people. The team is made up of researchers and engineers from top universities and leading AI labs, and it’s growing.

The Role

This is a model-training role, not a general ML role or a robotics infrastructure role. The role involves fine-tuning LLMs so that they can operate robots, with models that have to run in real time. The ideal candidate trains these models and understands their behavior when they fail on real hardware. Successful candidates know both how to train a model and how a robot behaves.

Key Responsibilities

• Fine-tune and post-train LLMs and VLA (vision-language-action) models for real-time robotic control.
• Run post-training experiments and evaluate foundation models on robotics tasks.
• Debug model behaviour on robots, identifying why a model behaves as it does and whether the cause is the model, the environment, or the setup.
• Work with researchers and founders to turn training results into reliable robot capability, comparing simulation with real-robot evaluation.

Requirements & Tech Stack

• Strong Python and engineering fundamentals.
• Careful experimental judgment: ability to determine if an improvement is real and understand limitations of evidence.

Qualifications

• Hands-on experience training or post-training models.
• Proficient with robotics and embodied AI.
• LLM experience and RL applied to robotics are relevant.
• Experience with VLA models, world models, or world action models is ideal.
• Recent graduates with substantial research or project work are welcome.
• A Master’s or PhD is preferred.