Research Scientist - Interactive Avatars
Synthesia · Remote · Europe · London · Munich · Zurich ·
COMPENSATION
€•• – ••k
Listed on the employer's job board (Ashby)
- Fully remote
- Generative AI focus
- Video diffusion models
- Senior research role
- PyTorch environment
Join the Interactive Avatars team at Synthesia to advance the state of AI video agents. You will build models that understand user audio and video inputs to generate natural, real-time conversational reactions including gaze and body language. The team transforms breakthrough research into shipped product capabilities for enterprise communication.
What you'll do
- Own scoped research problems for dyadic interaction modeling end to end
- Advance perceptual layers for understanding user audio and video inputs
- Post-train multimodal models to generate natural dyadic interactions
- Adapt diffusion models to conditioning signals like turn-taking and cues
- Build robust evaluation frameworks to track interaction quality continuously
- Partner with data teams to define needs and shape high-quality datasets
- Run rigorous experiments and share findings to inform technical decisions
What you'll need
- Strong ML background with hands-on experience in diffusion models
- Publications at top-tier venues like CVPR, NeurIPS, or ICLR
- Experience taking research ideas through to working implementations
- Proficiency in PyTorch and modern ML tooling for large-scale training
- Ability to clearly communicate hypotheses, experiments, and results
Nice to have
- Experience with real-time or streaming generation techniques
- Knowledge of distillation for low-latency inference
- Background in audio-driven facial or full-body motion modeling
- Expertise in conversational modeling like backchanneling
- Experience mentoring students or junior researchers
Stack
- PyTorch
- Diffusion Models
- Multimodal Models
- Generative AI
How you'll work
- Fully remote position
- Part of 40+ person R&D department
- Fast-paced environment with high ownership
What you get
- Work on production-scale video foundation models
- High-ownership environment where work ships to product
- Influence direction of next-generation synthetic human technology