Open position
Diffusion training and fine-tuning
Train and fine-tune the image and video models our API serves. A new checkpoint ships when more of its outputs pass our acceptance checks, not when it wins a benchmark.
What you'll do
- Fine-tune image and video diffusion models for the jobs customers run most, such as presenters, products and image-to-video.
- Build the data and evaluation loops that tell a better checkpoint from a lucky one.
- Turn research results into models the serving team can run at our prices.
- Write up what worked and what did not, so the next run starts further ahead.
What you bring
- Hands-on experience training or fine-tuning diffusion or other generative models.
- Strong PyTorch, and comfort with distributed training on multi-GPU machines.
- A habit of measuring before claiming.
Good to have
- Work on video models, adapter training or distillation.
- Papers, open-source code or demos you can show us.