VLA/WAM Research Engineer - Multimodal Robot Foundations

Flexion; San Francisco (CA); On-site; USD 120,000 - 190,000; Full time

Benefits offered by this job

Competitive compensation

Relocation & permit sponsorship

Central Zürich office

Energetic, collaborative culture

Enhanced pension plan

Job summary

Flexion is advancing the intelligence layer powering humanoid robots by developing foundation models, VLA/WAM architectures, and training pipelines over diverse embodied data. The team focuses on big-model training, adaptation, and continual improvement from real and simulated robot experiences.

You will contribute to research across model architecture, pretraining, and post-training, collaborating with a world-class team to push the boundaries of robotic intelligence and deployment.

Qualifications

• Strong Python and PyTorch skills with hands-on experience in building and training modern deep-learning models.
• Experience with VLA/VLM architectures, multimodal foundation models, world models, large-scale pretraining, robot learning, or RL/post-training.
• Ability to design rigorous experiments and evaluate model performance across tasks and embodiments.

Responsibilities

• VLA/WAM Architecture: Design and develop multimodal foundation models for general-purpose robotic control.
• Foundation Model Pretraining: Lead large-scale pretraining across diverse robot embodiments and data sources.
• Training Recipes & Scaling: Improve data mixtures, objectives, optimization, and distributed training workflows.

Skills

Python

PyTorch

Deep learning

Tools

PyTorch

Job description

Flexion is advancing the intelligence layer powering humanoid robots by developing foundation models, VLA/WAM architectures, and training pipelines over diverse embodied data. The team focuses on big-model training, adaptation, and continual improvement from real and simulated robot experiences.

You will contribute to research across model architecture, pretraining, and post-training, collaborating with a world-class team to push the boundaries of robotic intelligence and deployment.