P
Video Research Engineer (Remote — Worldwide)
FeaturedPyTorchGenerative AIDiffusion ModelsVideo SynthesisDistributed TrainingCUDAComputer VisionLLMsPythonScalable Infrastructure
About the role
Join Pika's core research team to push the boundaries of what is possible in generative cinema and motion synthesis. As a Video Research Engineer, you will design and train large-scale diffusion models that power our flagship video generation platform, directly impacting millions of creators worldwide. You will work in a high-velocity, remote-first environment to bridge the gap between academic breakthroughs and production-ready creative tools.
Responsibilities
- Develop and implement novel architectures for high-fidelity video synthesis and temporal modeling.
- Scale generative model training across massive GPU clusters, optimizing for both throughput and memory efficiency.
- Research and apply advanced techniques in flow matching, latent video diffusion, and controllable generation (ControlNet, IP-Adapter).
- Collaborate with the product team to translate research milestones into user-facing features like motion control and style transfer.
- Stay at the forefront of AI research, prototyping new papers and identifying opportunities for state-of-the-art improvements.
- Build and maintain robust evaluation benchmarks for video quality, motion fluidity, and prompt adherence.
Requirements
- Proven track record of training large-scale generative models (Diffusion, Transformers, or GANs) at scale.
- Deep expertise in PyTorch and distributed training frameworks such as DeepSpeed, FSDP, or Megatron-LM.
- Strong understanding of video-specific challenges, including temporal consistency, 3D motion dynamics, and long-range dependency modeling.
- Experience managing large-scale datasets, including data curation, filtering pipelines, and multimodal alignment (CLIP/CLAP).
- History of high-quality research output, evidenced by top-tier conference publications (CVPR, ICCV, NeurIPS) or significant open-source contributions.
- Ability to thrive in a fast-paced startup environment with minimal oversight and a high degree of technical autonomy.
- Excellence in writing clean, maintainable, and highly optimized CUDA or Python code.
Benefits
Equity, top compensation, compute access.
Disclaimer: MMagic.ai connects talented people with AI companies around the world. While we work hard to feature quality opportunities, we don't independently verify employers, candidates, salaries, or hiring outcomes. We encourage you to research each opportunity and company before applying or making an offer.