Full-time position
Research Scientist: Energy Based Models
Astera Institute
- Emeryville, United States · Hybrid
- Artificial Intelligence, Machine Learning and
About
Astera’s AI Research group pursues brain-inspired foundations for general intelligence, on the view that understanding the brain’s computational principles is a path to more capable machines. This research scientist develops and implements new architectures and learning and inference algorithms for neuroscience-inspired energy-based world models built on hierarchical latent-variable models, with a focus on representations that allow parallel distributed planning, and designs experiments testing hypotheses about high-level cognition, causal representations, sensory-motor integration, active inference and unsupervised learning.
The team works alongside neuroscientists and software engineers and publishes openly. The posted Bay Area salary range is $150,000–$300,000 USD.
Requirements
Preferred: PhD in computer science, electrical engineering, neuroscience, physics or a related quantitative field (exceptional candidates with non-traditional backgrounds are considered); strong foundations in deep learning, graphical models and information theory; expert-level skills in PyTorch or JAX with clean, reproducible research code; interest in robustness, generalisation and common-sense reasoning; interest in the computational and architectural principles of the mammalian brain.
Topics
- Energy Based Models, search World Wide
- world models, search World Wide
- active inference, search World Wide
- planning, search World Wide
- hierarchical latent-variable models, search World Wide
- parallel distributed planning, search World Wide
- high-level cognition, search World Wide
- causal representations, search World Wide
- sensory-motor integration, search World Wide
- unsupervised learning, search World Wide
Source: Astera Institute