Hedra
Research Engineer
About this role
Hedra is seeking a Research Engineer to develop and deploy action-conditioned world models and vision-language-action systems that bridge generative AI with real-world physical applications. You'll lead pre-training and post-training efforts, collaborate with industrial partners, and contribute publishable research at the intersection of generative modeling and robotics.
What you'll do
- Design and implement pre-training and post-training pipelines for action-conditioned world models and VLA systems
- Develop training methodologies including fine-tuning, reinforcement learning, and large-scale multimodal learning
- Create training and evaluation datasets from simulation with domain randomization and sim-to-real transfer strategies
- Build distributed training infrastructure using PyTorch, FSDP, and DeepSpeed
- Process multimodal data pipelines integrating video, sensory inputs, and action sequences
- Partner with industrial clients to adapt models for real-world physical AI deployment
What they're looking for
- Pre-training and post-training of large generative models
- PyTorch and distributed training frameworks (FSDP, DeepSpeed)
- Machine learning optimization and large-scale data processing
- Vision-language models (VLMs) and world models
- Video understanding and temporal reasoning
- Robotics or embodied AI systems
- Sim-to-real transfer techniques
- Multimodal learning architectures
Benefits
- Competitive compensation and equity
- Health insurance (Medical PPO, Vision, Dental)
- 401(k) retirement plan
- Office lunch and snacks
Opens the application — the Jobs AI extension fills it for you. Set up autofill
Opens the official application on the employer’s site. No login required.
Hedra
Hedra develops action-conditioned world models and vision-language-action systems that apply generative AI to real-world physical applications. The company is hiring Research Engineers to lead pre-training and post-training efforts, collaborate with industrial partners, and advance research at the intersection of generative modeling and robotics.
View all jobs at HedraLikely interview questions
- Walk us through a project where you implemented pre-training or post-training for a large generative model. What was the most challenging part of the pipeline?
- Describe your experience with distributed training frameworks like FSDP or DeepSpeed. How did you handle communication bottlenecks or memory constraints?