Helix AI Engineer, Video Pretraining
At a glance
Mid-level AI Research role at Figure. San Jose · $200,000–$400,000 base.
aimlroles summary, based on the employer's posting.
What you'll do
- Train video foundation models using internet-scale footage and recordings collected by robots
- Shape pretraining so models learn motion, timing and object interactions from raw video
- Build reusable representations for perception, tracking, prediction and control
- Test transformer and diffusion approaches, then measure temporal understanding and generalization
- Improve video data pipelines and distributed training for throughput, compute and memory efficiency
What you bring
- You have trained large models on video or other high-dimensional sequential data
- You understand deep learning methods for video, vision or multimodal systems and large-scale pretraining
- You can build in Python with PyTorch and work across distributed training systems and large GPU clusters
- You bring strong experimental practice and software engineering skills for reliable, scalable research systems
Who this fits
This role suits an engineer who can independently set research direction, test model designs quickly and work across research and engineering. It is based in San Jose, and the stated base salary range is $200,000 to $400,000 per year.
From the employer
Figure is an AI robotics company developing autonomous general-purpose humanoid robots. The goal of the company is to ship humanoid robots with human level intelligence. Its robots are engineered to perform a variety of tasks in the home and commercial markets. Figure is headquartered in San Jose, CA.
Our Helix team is responsible for developing the core AI systems that power humanoid autonomy. We are looking for a Helix AI Engineer, Video Pretraining to lead the development of large-scale video foundation models trained on diverse real-world and robot-collected data.
This role focuses on pretraining models that learn from raw video—capturing motion, interaction, and temporal structure—to enable downstream capabilities in perception, prediction, and embodied reasoning.
Responsibilities:
-
Design and train large-scale video foundation models on diverse datasets spanning internet-scale video and robot-collected data.
-
Develop pretraining strategies that capture temporal dynamics, motion, and object interaction from raw video sequences.
-
Build models that learn transferable representations for downstream tasks such as perception, tracking, prediction, and control.
-
Explore architectures for video understanding and generation, including transformer-based and diffusion-based approaches.
-
Implement efficient data pipelines and training strategies for high-throughput video ingestion and large-scale distributed training.
-
Optimize model performance across compute, memory, and training efficiency constraints.
-
Collaborate closely with generative modeling, agent, and robot learning teams to integrate pretrained models into the autonomy stack.
-
Design evaluation frameworks and benchmarks to measure temporal understanding, prediction quality, and generalization.
Requirements:
-
Experience training large-scale models on video data or other high-dimensional sequential modalities.
-
Strong understanding of modern deep learning architectures for video, vision, or multimodal systems.
-
Experience with large-scale pretraining, including dataset curation, training dynamics, and scaling laws.
-
Proficiency in Python and deep learning frameworks such as PyTorch.
-
Experience working with distributed training systems and large GPU clusters.
-
Strong experimental rigor and ability to iterate quickly on model design and training strategies.
-
Solid software engineering skills and ability to build scalable, reliable systems.
-
Ability to operate independently and drive ambiguous, high-impact research directions.
Bonus Qualifications:
-
Experience working on frontier video models or multimodal foundation models.
-
Background in video diffusion, autoregressive video modeling, or world models.
-
Experience at leading AI labs such as OpenAI, Google DeepMind, Google, ByteDance, Midjourney, or Adobe.
-
Experience with large-scale dataset construction and filtering for video pretraining.
-
Familiarity with robotics, embodied AI, or learning from egocentric / first-person video.
-
Publication record in machine learning, computer vision, or multimodal AI.
The US base salary range for this full-time position is $200,000 to $400,000 per year.
The pay offered for this position may vary based on several individual factors, including job-related knowledge, skills, and experience. The total compensation package may also include additional components/benefits depending on the specific role. This information will be shared if an employment offer is extended.