Staff Research Scientist, Exotic AI
Build next-generation training infrastructure for physical AI models that perceive, reason, and act in structured environments. Lead development of representation models, latent world models, and policy optimization systems.
About the job
Responsibilities
- Design and build scalable training infrastructure for representation models (e.g., contrastive and self-supervised approaches like CLIP/SigLIP, DINO/MAE, and joint-embedding predictive architectures)
- Develop latent world models that learn environment dynamics through imagined rollouts, enabling model-based reasoning and planning (Dreamer-style, I-JEPA/V-JEPA families)
- Architect and implement action/policy model pipelines, including vision-language-action models and diffusion-based policy learning
- Build generative simulator frameworks that produce controllable, physically plausible future states (video world models in the spirit of Cosmos/Genie/Sora)
- Develop multimodal generative model capabilities that fuse visual, language, and structured inputs for downstream reasoning and decision-making
- Lead cross-team technical decisions on training frameworks, data pipelines, and model evaluation infrastructure
- Drive research-to-production pathways, translating prototype systems into reliable, performant platform capabilities
- Contribute to the broader research community through publications, open-source releases, and collaboration with academic partners
Requirements
- 8+ years of relevant experience in machine learning engineering, AI research, or a closely related field (or equivalent experience)
- Deep expertise in at least two of the following: representation learning, world models, reinforcement learning, generative modeling, robotics/embodied AI, or scientific ML
- Hands-on experience training large-scale models (vision, language, or multimodal) with distributed compute
- Strong software engineering fundamentals: system design, performance optimization, and production-quality code
- Demonstrated ability to drive cross-team technical initiatives with ambiguity and limited direction
- Track record of translating research ideas into working systems at scale
- MS or Ph.D. in Computer Science, Machine Learning, Robotics, Physics, or a related field, or equivalent experience
Nice-to-Haves
- Experience with latent dynamics modeling, model-based RL, or physics-informed neural networks (GraphCast, FourCastNet, AlphaFold-style architectures)
- Contributions to open-source ML frameworks or foundation model training codebases
- Background in scientific/structured models (molecular modeling, materials science, weather/climate)
- Experience building controllable video generation or neural simulation environments
- Publications at top venues (NeurIPS, ICML, ICLR, CVPR, CoRL, RSS)
Skills
PyTorch, TensorFlow, JAX, Representation Learning, World Models, Reinforcement Learning, Generative Modeling, Robotics, Vision-Language Models, Diffusion Models, Distributed Training, Model Evaluation
Similar jobs
AI Research jobs
Shield AI
Washington, DCLeads technical direction and develops maritime autonomy capabilities for unmanned surface and underwater vehicles, including motion planning, localization, safe behaviors, and heterogeneous multi-agent collaboration. Requires deep robotics and unmanned-systems experience, strong C++/Python skills, and senior technical leadership.
Shield AI
San Mateo, CA
Leads design, implementation, integration, and field validation of tactical autonomy software for unmanned systems and multi-agent missions. Requires extensive autonomy or robotics experience, strong C++ and Python skills, and the ability to obtain a SECRET clearance.
Upstart
United States
Evaluates model and Generative AI risks across Upstart Bank’s model inventory, conducting risk assessments, monitoring reviews, quantitative analyses, and governance activities. Requires a quantitative master’s degree, 4+ years of relevant experience, and coding skills in Python, R, or similar languages.
Anthropic
San Francisco, CA
Research and evaluate frontier AI capabilities for cybersecurity, rapidly prototyping tools, designing rigorous benchmarks, and helping operationalize reliable capabilities into products. Requires deep security expertise, strong technical communication, and at least seven years of relevant experience.
Order.co
United States
Own the architecture, delivery, evaluation, and production operations of AI capabilities embedded in procurement and finance workflows. The role requires 10+ years in applied AI or machine learning, deep LLM and agent expertise, and experience delivering measurable production outcomes.