Machine Learning Engineer

Machine learning for humanoid robots: imitation learning, RL, VLAs and world models, trained and deployed on real hardware.

Wegens grote belangstelling verzoeken wij u snel te solliciteren als u een goede match bent voor deze functie.

At the Humanoid Application Center (HAC) we believe humanoid robotics will become a flexible, universal automation layer that fundamentally changes how physical work gets done. We work with partner companies on real-world use cases, and are building the team and internal software stack to enable this. All of it happens on our own floor of humanoids in Schiedam (Rotterdam area).

You’ll be the ML specialist on a small, cross-disciplinary engineering team. The core challenge in humanoids is closing the gap between impressive demos and reliable deployment in real work environments. You work on the learning side of that gap.

What you’ll do

• 
  
• Train, fine-tune and deploy models across the range: small embedded models up to foundation models, VLMs and world models
  
• Build and own our learning loops: teleop data collection, dataset curation, training pipelines, evaluation, redeployment
  
• Develop imitation learning and reinforcement learning policies for manipulation and locomotion, in simulation and on hardware
  
• Build out our simulation environments (Isaac Sim/Lab, MuJoCo or similar) for training, evaluation and sim-to-real transfer
  
• Take partner use cases from "not yet possible" to demonstrated and deployed
  

What you bring

• 
  
• Hands‑on experience training deep learning models yourself, not only using them via APIs
  
• Solid Python and PyTorch (or JAX)
  
• Experience deploying models to embedded hardware (quantization, TensorRT, ONNX or similar)
  
• MSc or PhD in ML, AI, robotics or CS, or equivalent demonstrated experience
  
xgiwjmb
• Strong pluses: RL or IL on real robots, sim-to-real, ROS 2, VLAs, world models, diffusion policies, MLOps practice
  

What we offer

• 
  
• A working floor of state‑of‑the‑art humanoids: train a model in the morning, run it on hardware in the afternoon
  
• Real industrial use cases with committed partners, and founding‑stage influence on the stack
  
• On‑site role in Schiedam (Rotterdam area) with hybrid options
  

We value complementary backgrounds and encourage you to apply even if you don't tick every box.