Helix AI Engineer Pretraining
Figure is an AI robotics company developing general-purpose humanoid robots and its Helix vision-language-action AI system.
Funding history
About Figure
Figure develops and deploys general-purpose humanoid robots for commercial and household tasks. Its current Figure 03 robot is powered by Helix, an onboard generalist vision-language-action model; the company also operates the Index data-collection service.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will design and train foundation models using text, vision, and robot data. You will develop pretraining strategies, optimize distributed GPU training, study scaling laws and dataset mixtures, evaluate model performance, and contribute to fine-tuning, alignment, and adaptation.
Requirements
- Foundation-model training or pretraining for LLMs or multimodal systems
- Deep learning architecture
- Transformer
- Distributed training
- Optimization
- Python
- PyTorch
- Software engineering
- Experimental rigor
Responsibilities
- Design and train large-scale multimodal foundation models
- Develop pretraining strategies for generalization, reasoning, and transfer
- Implement transformer-based and emerging foundation-model architectures
- Work on scaling laws, dataset mixtures, and training dynamics
- Build distributed training pipelines across multi-node GPU clusters
- Integrate pretrained models into the autonomy stack
- Design evaluation frameworks
- Contribute to fine-tuning, alignment, and model adaptation