SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Synthesia is a $4B AI video platform used by over 90% of the Fortune 100. The company is seeking a Research Scientist to join a 40+ person R&D team focused on generative AI, specifically avatar-centric interactive video diffusion models.
You will own well-scoped research problems end-to-end, advancing the state of the art in interactive agents that listen, respond, and react naturally in real-time conversations. Your work will span understanding user audio and video inputs, generating contextually appropriate reactions, and post-training multimodal models for rich dyadic interactions. You'll adapt diffusion models to new conditioning signals (conversational state, turn-taking, listener cues), build robust evaluation frameworks for interaction quality, partner with the data team on dataset definition, and run rigorous experiments to inform technical decisions.
Synthesia emphasizes a culture of building and shipping—your research will move from breakthrough ideas to production capabilities. The company is backed by premier investors including Accel, Nvidia Ventures, Kleiner Perkins, and GV, with total funding exceeding $530 million.
REQUIREMENTS:
- Strong ML background with hands-on experience in diffusion models, ideally for video or avatar generation
- Publications at top-tier venues (CVPR, ICCV, ECCV, NeurIPS, ICML, ICLR, SIGGRAPH) in areas such as world models, dyadic interaction, or video diffusion, or equivalent demonstrated impact
- Proven ability to take research ideas through to working implementations with motivation to see them reach production
- Proficiency in PyTorch and modern ML tooling for large-scale training
- Clear communication of hypotheses, experiments, and results
NICE TO HAVE:
- Experience with real-time or streaming generation, including autoregressive video diffusion
- Distillation or other techniques for low-latency inference
- Audio-driven facial, gesture, or full-body motion modeling
- Conversational modeling (turn-taking, backchanneling, listener-response generation)
- Experience mentoring students or junior researchers
