Skip to main content

Bio

Sreejan Kumar, PhD

2025 Leon Levy Scholar in Neuroscience

Columbia University, New York University

Sub-disciplinary Category

Computational Neuroscience

Previous Positions

  • BS, Yale University
  • PhD, Princeton University (Dr. Tom Griffiths, Dr. Jon Cohen)

Bio

Sreejan Kumar obtained his BS with a dual major in Computer Science and Statistics & Data Science at Yale University, where he worked as a research assistant in the Department of Psychology. He then pursued a PhD at Princeton University with Professors Tom Griffiths and Jon Cohen. His research used methods from computational cognitive science to study flexible behavior and generalization in human and artificial intelligence. During graduate school, Sreejan’s research was recognized with accolades such as the Google PhD Fellowship and a NeurIPS Outstanding Paper Award. He is now an incoming postdoctoral fellow at the Columbia University Zuckerman Mind Brain Behavior Institute, working with Professor Lea Duncker, as well as a visiting academic at the NYU Department of Psychology, working with Professor Marcelo Mattar. His postdoctoral research focuses on modeling the neural mechanisms underlying flexible behavior, bridging cognitive theory with systems neuroscience.

Research Summary

Computational modeling of the neural mechanism by which the brain implements flexible, goal-directed behavior.

Technical Overview

Flexible, goal-directed behavior—where actions are flexibly and intentionally chosen to achieve specific outcomes—contrasts with habitual behavior, which is driven by routine and automaticity. These behavioral modes are thought to arise from complex loops that include the cortex and distinct subregions of the striatum: the dorsolateral striatum (DLS), associated with habitual motor learning, and the dorsomedial striatum (DMS), implicated in flexible control. Sreejan develops recurrent neural network models to study how these corticostriatal circuits support different forms of behavior. Building on established models of DLS-driven motor actions, his work investigates how the DMS enables mental actions that reflect internal cognitive processes—such as gating information into working memory, accumulating evidence before making a decision, or planning towards a high-level goal. By building these models in tandem with analyzing neural recordings from both humans and animals, his research aims to uncover the general neurocomputational principles that support flexible behavior across species.