Pranav Misra, Ph.D.

AI · Language · Computational Neuroscience

Postdoctoral Fellow at the Center for Brain Science, Harvard University.

Pranav Misra

I am a postdoctoral fellow at the Center for Brain Science at Harvard University. I work on the geometry and structure of language representations in modern neural networks. I study how concepts, categories, and compositional structure are organized in high-dimensional spaces. I try to understand what makes some representations stable, interpretable and generalizable.

I completed my PhD under Professor Gabriel Kreiman at Harvard. For my dissertation, I built and analyzed what remains the largest multimodal intracranial dataset for language to date, covering phrases, sentences, images and naturalistic meaning. That work revealed a cortical region in the left orbital H-shaped sulcus, sitting above the left eye, that reliably encodes parts of speech and basic syntactic structure across individuals. I am now a postdoc with Professor Haim Sompolinsky. His theoretical neuroscience and statistical physics background shapes how I think about representation, structure and computation in both biological and artificial systems.

My undergraduate training was at BITS Pilani, where I majored in computer science. Some of my other favorite courses were about the theory of general relativity in physics and the analysis of generative grammars in linguistics. As an undergrad, I built an augmented-reality system for safer surgical planning and earned a GE Foundation Fellowship. That experience taught me how engineering ambition and real-world application can work alongside rigorous thinking. What drives me is curiosity. I like to understand how complex systems work and process information. That curiosity pushes me to use theoretical tools from computer science and physics to study language and natural intelligence.

Invited Talks

  • Genentech R&D (gRED), August 2024
  • Cleveland Clinic Neurosurgery Grand Rounds, November 2023
  • LMU–Harvard Young Scientists Forum, April 2023
  • Center for Brain Science Seminar, February 2023
  • Brains and Behavior in the Wild, November 2022

Publications

Invariant neural representation of parts of speech in the human brain
Invariant neural representation of parts of speech in the human brain
Misra, P., Shih, Y.-C., Yu, H.-Y., Weisholtz, D., Madsen, J. R., Sceillig, S. & Kreiman, G. (2024). bioRxiv preprint.
Multimodal syntactic and semantic processing of sentences in the human brain
Multimodal syntactic and semantic processing of sentences in the human brain
Misra, P., Sharma, A., Shih, Y.-C., Yu, H.-Y., Weisholtz, D., Madsen, J. R., Stone, S., Bulacio, J., Bingaman, W. & Kreiman, G. Manuscript in preparation (2024).
Brain TreeBank: Intracranial recordings of neural activity with multimodal stimulus
Brain TreeBank: Intracranial recordings of neural activity with multimodal stimulus
Wang, C., Yaari, A., Singh, A., Subramaniam, V., Rosenfarb, D., DeWitt, J., Misra, P., Madsen, J., Stone, S., Kreiman, G., Katz, B., Cases, I. & Barbu, A. (NeurIPS, 2024).
Neural processing of nouns and verbs with large-scale intracranial recordings
Neural processing of nouns and verbs with large-scale intracranial recordings from naturalistic language
Wang, C., Yaari, A., Singh, A.K., Cases, I., Subramaniam, V., Rosenfarb, D., Misra, P., Madsen, J.R., Stone, S., Kreiman, G., Katz, B. & Barbu, A. (2024, manuscript).
Hierarchical Sequence Composition in Recurrent Networks with Finite Automata
Hierarchical Sequence Composition in Recurrent Networks with Finite Automata
Misra, P. (2024, manuscript).
Minimal memory for details in real life events
Minimal memory for details in real life events
Misra, P., Marconi, A., Peterson, M. & Kreiman, G. (2018). Nature Scientific Reports.