Beneficial RL
Training models for broadly useful behavior that persists under pressure.
OpenAI
Akshay Jagadeesh
I'm a Research Scientist at OpenAI working to build AI systems that are deeply and persistently beneficial, with a particular focus on health and medicine. Previously, I was a computational neuroscientist studying how human and primate brains represent visual objects.
Selected work
View allAI for health
Exploring how AI can accelerate progress in health, medicine, and scientific discovery.
OpenAI
Neural computation
Using models to connect neural activity with perception and behavior.
Stanford · Harvard
Vision and models
Studying where biological and artificial visual systems align and diverge.
Published in PNAS, Nature Communications, NeurIPS workshop, ICLR workshop, CCN
Publications
View allReinforcement Learning Towards Broadly and Persistently Beneficial ModelsOpenAI Alignment Blog · 2026HealthBench Professional: Evaluating Large Language Models on Real Clinician ChatsOpenAI · 2026Assessing equivariance in visual neural representationsCCN · 2024Texture bias in primate ventral visual cortex and deep neural network models of visionICLR ReAlign · 2024Texture-like representation of objects in human visual cortexPNAS · 2022