Applied Research Engineer (Video Generative AI)

Location: Europe; remote

Company

Synthesia is an AI video platform for businesses that develops generative AI products for visual communication and enterprise skill development.

What you will do

• Develop and scale latent video diffusion models for human-centric video generation.
• Design conditioning mechanisms for pose, emotion, script, and camera control while preserving visual fidelity.
• Advance distributed training strategies using DDP, FSDP, DeepSpeed, and sequence parallelism at multi-node scale.
• Improve training stability and build evaluation frameworks combining automated metrics with structured human evaluation.
• Optimize inference for low latency, high resolution, and cost efficiency in production.
• Run controlled ablations and maintain reproducible engineering practices, including experiment tracking, CI/CD, and monitoring.

Requirements

• Strong experience training deep learning models at scale, including large-scale multi-GPU and multi-node training.
• Strong Python and PyTorch skills, with experience using CUDA.
• Hands-on experience with diffusion models; image-domain experience is required and video experience is preferred.
• Good understanding of distributed training and inference with DDP, FSDP, DeepSpeed, or similar technologies.
• Ability to design controlled experiments and interpret noisy results.
• Experience optimizing inference systems for production.

Nice to have

• Experience with video diffusion models and avatar or human-centric generation.
• Familiarity with world or interactive models.
• Experience with GANs or VAEs.
• Experience with AWS, SLURM, Docker, GitHub, and CI/CD pipelines.

Culture & Benefits

• Work on production-scale video foundation models in a generative AI company.
• Contribute to human-centric video generation used by businesses worldwide.
• Work in a high-ownership environment focused on shipping practical outcomes.
• Collaborate with AI researchers and engineers on AI safety, ethics, and security.