VLA/WAM Research Engineer: Multimodal Robots

Flexion Robotics; San Francisco (CA); On-site; USD 150,000 - 230,000; Full time

Benefits offered by this job

Competitive compensation

Enhanced pension plan

Enhanced holiday & paid leave perks

Relocation & permit sponsorship

Central Zürich office with top-tier… (

Job summary

Flexion Robotics is building the intelligence layer for humanoid robots. We seek a VLA/WAM Research Engineer to design and train scalable, multimodal foundation models that power robotic control and generalization across tasks and embodiments.

You will work across model architectures, pretraining, adaptation, and post-training, applying RL and continual learning to improve real-world robot performance. Strong Python/PyTorch skills are required.

Qualifications

• Experience with Python and PyTorch in deep learning projects.
• Depth in one or more areas: VLA/VLM architectures, multimodal models or world models.
• Familiarity with large-scale pretraining and continual learning concepts.

Responsibilities

• Design multimodal foundation models combining vision, language, robot state, actions, and context.
• Lead large-scale pretraining across diverse robot embodiments and data sources.
• Improve training recipes, objectives, and distributed workflows for scalable models.
• Develop world-action models for prediction, planning, and policy improvement.
• Create methods to adapt pretrained models to new tasks, environments, and embodiments.
• Advance reinforcement learning and post-training techniques using real and simulated data.
• Design rigorous experiments and evaluate models in simulation and on real robots.
• Demonstrate strong Python and PyTorch skills in all tasks.
• Own substantial model development, training experiments, and evaluation decisions.

Skills

Python

PyTorch

VLA/VLM architectures

Multimodal models

Large-scale pretraining

Robot learning

RL / post-training

Experiment design

Job description

Flexion Robotics is building the intelligence layer for humanoid robots. We seek a VLA/WAM Research Engineer to design and train scalable, multimodal foundation models that power robotic control and generalization across tasks and embodiments.

You will work across model architectures, pretraining, adaptation, and post-training, applying RL and continual learning to improve real-world robot performance. Strong Python/PyTorch skills are required.