Research Scientist World Models
SpAItialVisit SpAItial website
SpAItial is a London AI company developing physically grounded world models that generate and reason about explorable 3D environments.
London, United Kingdom
Funding history
About SpAItial
SpAItial develops the Echo model family and an app/API for generating, editing, exploring, sharing, and exporting persistent 3D Gaussian Splat worlds from text, images, and panoramas.
Skills
About the Role
You will architect generative world models that reason about space, time, and physics. You will design image and video diffusion models, train them at scale on distributed cloud infrastructure, maintain training data, create model demonstrations, and productionize and optimize models at test time.
Requirements
- PhD in computer science or a related field
- Top-tier publication record in machine learning, vision, or graphics
- Knowledge of image or video diffusion generative models
- Knowledge of diffusion-transformer architectures including DiT
- Experience with deep-learning frameworks including PyTorch
- Expert coding skills and ability to rapidly iterate on machine-learning experiments
- Experience using modern AI coding tools
Responsibilities
- Architect generative world models that reason about space, time, and physics
- Design and develop image and video diffusion foundational models
- Train models at scale on distributed cloud infrastructure
- Develop demonstrations for trained model prototypes
- Process and maintain large datasets for model training
- Productionize models and optimize them at test time