What are the responsibilities and job description for the Research Scientist - Diffusion Models position at kadence?
Founding Research Scientist – Generative Video
📍 Seattle, WA (Onsite, 5 days/week) | 💼 Full-time
About the Founders
Founded by a collective of PhD researchers from leading global universities (Apple, Stanford & Oxford PhD Alumni) with backgrounds at top AI labs and major technology companies, combining deep theoretical insight with proven large-scale product execution.
About the Company
Our client is a stealth-mode AI startup creating breakthrough technology at the intersection of perception, learning, and expression. The team’s mission is to build intelligent systems that can observe and interpret subtle human cues — visual, vocal, and behavioral — to enable natural, emotionally aware digital interaction.
They’re developing foundational models that unify multiple input streams to make communication between humans and machines feel effortless and authentic.
Role Overview
This position is ideal for a scientist who thrives at the boundary of research and product impact. As the Founding Research Scientist for Generative Video, you’ll lead experimentation on advanced video generation methods, designing architectures that model motion, timing, and realism at scale.
You’ll have freedom to explore new diffusion-based approaches, define metrics for visual fidelity and consistency, and bring early-stage research into practical real-time applications.
Key Responsibilities
- Design and implement generative video models that synthesize realistic movement, depth, and texture.
- Develop scalable pipelines for dataset creation, model training, and system evaluation.
- Explore diffusion, transformer, and temporal modeling techniques to enhance continuity and expression.
- Partner with engineers and scientists across modalities (speech, text, vision) to integrate outputs into unified frameworks.
- Evaluate model behavior, benchmark performance, and publish findings where appropriate.
- Transition promising prototypes into production-ready, low-latency systems.
Required Background
- PhD or comparable experience in Machine Learning, Computer Vision, or Graphics.
- Proven expertise in video or 3D generative modeling, preferably using diffusion or transformer architectures.
- Strong applied experience with PyTorch, JAX, or similar toolkits.
- Understanding of data curation, distributed training, and large-scale evaluation.
- Publications, patents, or open-source contributions demonstrating research excellence.
- Self-directed mindset with enthusiasm for rapid iteration and open-ended discovery.
What Makes This Unique
- Early technical ownership in shaping the research roadmap of a deep-tech organization.
- Opportunity to work on high-impact, unexplored problems in real-time generative media.
- Close collaboration with a multidisciplinary team of machine learning, vision, and graphics experts.
- Competitive compensation, strong equity participation, and top-tier backing from investors in applied AI.