Staff Machine Learning Engineer - Action Models
About this role
Atoms is building Physical AI and real-world robots for industries like food, mining, and transport to make complex physical operations more reliable, scalable, and productive. The team integrates hardware, software, AI, and operations to deploy machines into real environments that can understand, predict, and control the physical world.
As a foundational technical leader in the AI organization, this individual will focus on action models that enable intelligent machines to reason about their environment, make decisions, and translate those decisions into physical actions. The work connects modern foundation models and learned representations to action, moving beyond independently engineered components toward models capable of learning sophisticated behaviors.
Ideal candidates are deeply technical individual contributors who thrive in open-ended research spaces and can influence long-term AI architecture. The role requires navigating ambiguity, building robust systems that function outside research labs, and partnering across robotics, perception, and simulation disciplines.
What you will do
- Define and help build the technical architecture for action models and learned decision-making systems.
- Develop models and learned policies that translate physical world representations into decisions, plans, and actions.
- Research and apply techniques across reinforcement learning, imitation learning, and behavior learning.
- Explore vision-language-action and other multimodal architectures for physical AI applications.
- Train and evaluate models using large-scale real-world, simulated, and synthetic data.
- Partner with cross-functional teams spanning perception, world models, autonomy, and ML infrastructure.
- Provide technical leadership through research direction, architecture reviews, and hands-on engineering in Python or C++.
What you need
- Deep expertise in machine learning with experience developing models for robotics, decision-making, autonomous systems, or embodied intelligence.
- Strong understanding of modern deep learning architectures applied to physical systems and sequential decision-making.
- Experience with reinforcement learning, imitation learning, planning, control, robot learning, or embodied AI.
- Demonstrated ability to operate in ambiguous research spaces where the solution path is not yet defined.
- Strong software engineering fundamentals with hands-on coding ability in Python and/or C++.
- Track record of making consequential technical decisions and influencing broader research or engineering direction.
Benefits
- Medical, Dental, Vision, Disability, and Life Insurance
- Flexible Spending Account / Health Savings Account Options
- 401(k) and Equity
- Sick Time, Unlimited Flexible Time Off, and Paid Holidays
- Paid Parental Leave and Pre-Tax Commuter Benefit Plan
- Team lunch in the SoMa office every Tuesday and Thursday
Skills and tech stack
Questions candidates ask
What is the remote work policy?
This role is based in the San Francisco office. All office-based teams work onsite, five days a week.
What is the salary range for this position?
The base salary range for this role is $273,000 to $321,000 per year. Additional compensation may include equity awards, depending on individual qualifications.
What technologies and methods will I be working with?
You will work with machine learning architectures for decision-making, reinforcement learning, imitation learning, and multimodal architectures using Python and C++. The role involves training and evaluating models across real-world, simulated, and synthetic datasets.
The application is on the official ATOMS Careers page careers page.
Apply on company site