Research Scientist – Interactive Avatars

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

📋 Description

• Become part of a team comprising over 40 researchers and engineers in R&D focusing on generative AI and avatar-centric interactive video diffusion models.

• Play a pivotal role in shaping research direction for dyadic interaction modeling while managing well-defined research projects from inception to completion.

• Enhance the perceptual layer of interactive agents by analyzing user audio and video to generate contextually relevant responses.

• Perform post-training on multimodal models to create natural dyadic interactions based on user audio and video inputs.

• Modify diffusion models to incorporate conditioning signals including conversational state, turn-taking, and listener cues.

• Develop evaluation frameworks and testing suites to monitor interaction quality.

• Collaborate with the data team to identify data requirements and cultivate high-quality datasets.

• Conduct thorough experiments and disseminate findings that guide technical decision-making.

⛳️ Requirements

• Solid machine learning foundation and practical experience with diffusion models, preferably in video or avatar generation.

• Publications in prestigious venues like CVPR, ICCV, ECCV, NeurIPS, ICML, ICLR, or SIGGRAPH regarding world models, dyadic interaction, or video diffusion, or equivalent demonstrated impact.

• Proven ability to transform research ideas into functional implementations.

• Expertise in PyTorch and contemporary ML tools for large-scale training.

• Ability to clearly articulate hypotheses, experimental designs, and results.

• Experience with real-time or streaming generation, including autoregressive video diffusion (preferred).

• Familiarity with techniques for low-latency inference, such as distillation (preferred).

• Experience in audio-driven facial, gesture, or full-body motion modeling (preferred).

• Knowledge of conversational modeling techniques, including turn-taking, backchanneling, or listener-response generation (preferred).

• Experience in mentoring students or junior researchers (preferred).

• Eligibility to work legally in the country of employment without visa sponsorship, as indicated in the application form.

🏝️ Benefits

• Develop production-scale video foundation models within a rapidly growing Generative AI organization.

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

• Address challenging issues related to scaling, stability, and controllability.

• Shape the future of next-generation synthetic human technology.

• Work in a highly technical environment that emphasizes ownership and enables delivery of impactful work.