Description
As a Staff Research Engineer, you will join a team of 40+ Researchers and Engineers within the R&D Department working on cutting-edge challenges in the Generative AI space, with a focus on creating highly realistic, emotional and life-like Synthetic humans through text-to-video.
Within the team, you'll have the opportunity to work on the applied side of our research efforts and directly impact our solutions that are used worldwide by over 55,000 businesses. If you are an expert in post-training Diffusion models for generative AI, this is your chance.
In this position, you'll join our Post-Training Team that works on generative video creation. You will make it possible to apply our foundational models to production use cases. This includes projects such as:
- Apply DPO (direct preference optimisation) to pre-trained models.
- Adapt models to extend their capabilities, for instance, by changing conditioning inputs.
- Build solutions for dubbing and evaluate the quality of lip-sync.
- Implement post-training optimisation techniques, such as quantisation, pruning and distillation, to improve the efficiency of diffusion models used in avatar generation.
- Analyse and address challenges related to model performance, ensuring high-quality output in avatar rendering.
- Stay updated with the latest research and advancements in diffusion models, adversarial networks and post-training optimisation methods.
You have a background in Computer Vision / Computer Science and 3+ years of industry experience. You have knowledge of recent advancements in post-training techniques. You have worked with generative models for images and/or videos (Diffusion/GAN) preferably in the avatar domain. You are interested in doing research, trying new things and pushing the boundaries, going beyond what’s already known. You have experience in using most modern frameworks for machine learning and deep learning. You have great coding skills in Python and you care about writing clean code. You have experience with SDLC tools (Git), preferably CI/CD.