Research Scientist - Simplex
Core
Developing a rigorous theory of latent internal structure in neural networks to understand intelligence, safety, and behavior, applying this to both artificial and biological systems.
Role type
Senior Research Scientist (AI Safety & Interpretability)
Builds
New theory of intelligence, unsupervised methods for belief geometry recovery, and experimental tools for monitoring and controlling AI systems.
Domain
Artificial Intelligence, Neuroscience, Physics, Information Theory
Deliverable
production ML models
Required skills
Rigorous mathematics, deep learning, dynamical systems, probability, information theory, experimental design, code implementation, self-directed research, clear communication
Preferred skills
PhD in Physics, CS, neuroscience, or math, extensive ML experience, experience in interpretability
Technologies
Transformers, LLMs, Bayesian inference, attention mechanisms, residual streams
Responsibilities
Derive and extend theoretical frameworks for neural network representations, build unsupervised methods to recover belief geometries in real LLMs, design and execute experiments to test theories on complex cognitive tasks, collaborate on red teaming efforts to stress-test the framework, analyze biological neural data to test applicability of the theory
Seniority
Senior, hands-on IC
