Machine Learning Engineer, Physical AI (Modeling)

Machine Learning Engineer, Physical AI (Modeling)

Type: Full-time
Compensation: Competitive base salary + meaningful equity
Confidential Search | Represented by Samson Rose

About the Company

Our client is a venture-backed AI startup building fully autonomous factories. Their vision is simple but ambitious: CAD goes in, a finished product comes out, and production scales through AI rather than headcount.

They're building robotic assembly stations powered by closed-loop neural policies, trained primarily on millions of synthetic simulation clips and refined through a smaller set of real-world demonstrations. The company already works with paying Fortune 500 customers and has access to production data from real manufacturing lines, giving its team a rare opportunity to develop and improve Physical AI systems in real-world environments.

Precision is everything. Unlike tasks where small deviations are acceptable, manufacturing requires accuracy and repeatability down to the millimeter. The team is tackling these challenges head-on, building AI models that can perform complex physical tasks reliably at scale.

The Role

This is a hands-on Machine Learning Engineering role focused on building the core models that power autonomous robotic assembly. You'll own the model lifecycle end to end, from defining data requirements and developing neural architectures to training, debugging, deploying, and validating models on real factory floors.

This isn't a research-only role or one focused on simply integrating existing models. The team is looking for an engineer who can translate cutting-edge ML research into production-ready systems, experiment with new approaches, and make independent technical decisions. You'll work across model architecture, reinforcement learning, synthetic data, evaluation, and sim-to-real transfer, with the autonomy to drive projects from initial ideas through deployment.

The team is small, fast-moving, and highly technical, offering significant ownership over the models and systems that will shape the future of manufacturing.

What You'll Need

• 5+ years of hands-on industry experience training, debugging, and deploying machine learning models in production.

• Experience building models that act in the physical world, such as robotics, robotic manipulation, autonomous vehicles, or other embodied AI systems.

• Deep knowledge of modern neural network and foundation model architectures, optimization techniques, and hands-on experience with diffusion models and multimodal architectures.

• Practical reinforcement learning experience with methods such as PPO, SAC, or GRPO, including reward modeling in production settings.

• A strong understanding of model evaluation, including evaluation metrics, A/B testing, and offline and online evaluation frameworks.

• The ability to read research papers, implement advanced techniques, and adapt them to solve real-world engineering challenges.

• Strong programming and ML engineering skills, with the ability to take models from experimentation to reliable production deployment.

• A BS, MS, or PhD in Computer Science, Robotics, Computer Graphics, or a related field, or equivalent practical experience.

Nice to Have

• Experience with sim-to-real transfer, synthetic data generation, or co-training on simulated and real-world data.

• Familiarity with vision-language-action (VLA) models, imitation learning, or manipulation policy learning.

• Experience across the full ML stack, from data pipelines and training infrastructure to deployment and production monitoring.

• Experience working on high-precision or contact-rich physical tasks, such as robotic assembly.

• Open-source contributions or published research in machine learning, robotics, or embodied AI.

• Experience at an early-stage startup or in a small team where engineers have significant ownership and move quickly.