Senior Applied Research Engineer – Video

This is a fully remote position, open to applicants in Europe.

📋 Description

• Develop the next generation of production-ready foundation models tailored for human-centric video creation.

• Take ownership and execute comprehensive research and engineering initiatives from initial hypothesis to tangible production outcomes.

• Create and enhance latent video diffusion models focused on human-centric video generation.

• Design conditioning mechanisms for elements such as pose, emotion, script, and camera control, while ensuring high fidelity.

• Enhance distributed training methodologies utilizing DDP, FSDP, DeepSpeed, and sequence parallelism.

• Boost training stability across multi-node environments.

• Establish evaluation frameworks that integrate automated metrics with structured human assessments.

• Optimize inference processes for low latency, high resolution, and cost-effectiveness.

• Conduct controlled ablation studies and experiments to inform modeling choices.

• Uphold engineering standards that ensure reproducibility, experiment tracking, CI/CD, and monitoring.

• Collaborate with various teams and present findings in a scientific manner.

⛳️ Requirements

• Extensive experience in training deep learning models at scale.

• Proficient in Python and PyTorch.

• Practical experience with diffusion models; familiarity with image-domain is required, while video experience is preferred.

• Proven experience with large-scale multi-GPU / multi-node training.

• Solid understanding of distributed training, including DDP, FSDP, DeepSpeed, or similar frameworks.

• Capable of designing controlled experiments and interpreting complex results.

• Experience with video diffusion models (preferred but not mandatory).

• Background in avatar or human-centric generation (preferred but not mandatory).

• Knowledge of world or interactive models (preferred but not mandatory).

• Familiarity with GANs or VAEs (preferred but not mandatory).

• Experience in optimizing inference systems for production environments.

• Skills in Python, PyTorch, and CUDA.

• Familiar with DeepSpeed, distributed training and inference, sequence parallelism, AWS, SLURM, Docker, GitHub, and CI/CD pipelines.

• Ability to communicate effectively and present findings scientifically.

• Capable of working independently while actively collaborating across teams.

• Authorized to work in the country of employment without the need for visa sponsorship.

• No ongoing employer support necessary to maintain the right to work.

🏝️ Benefits

• Work remotely within Europe.

• Opportunity to contribute to production-scale video foundation models.

• Direct impact on products used by tens of thousands of businesses globally.

• Engage in human-centric video generation with significant real-world implications.

• Join a highly technical, high-ownership environment where your contributions are valued.