Open to workConsidering industry research positions starting fall 2026.
DOCX vCard Plain text

Dr. Arun Patel

Computational Neuroscientist

I build models that decode what populations of neurons are computing. Bridging Bayesian inference, deep learning, and experimental neuroscience.

Cambridge, MA arunpatel.science Google ScholarORCIDGitHub

Research Summary

I develop computational methods for understanding how neural circuits implement decision-making and memory. My work combines large-scale electrophysiology (Neuropixels, calcium imaging) with probabilistic modeling and representation learning. During my postdoc at MIT, I introduced a variational framework for latent-variable models of neural population dynamics that has been adopted by six labs worldwide. I have published 14 peer-reviewed papers (h-index 12, 840+ citations) and maintain three open-source analysis libraries used across the field.

Research Experience

Postdoctoral Research Associate

MIT — McGovern Institute for Brain Research

2022 – Present Cambridge, MA

Advisor: Prof. Mehrdad Jazayeri. Neural dynamics of timing and motor planning in primate prefrontal cortex.

  • Developed NeuroPLDI, a variational autoencoder for inferring latent dynamics from Neuropixels recordings across 400+ neurons simultaneously.
  • First-authored a Nature Neuroscience paper showing that prefrontal population trajectories encode elapsed time as a continuous manifold.
  • Built and released the pyLDI library (Python/JAX) — 600+ GitHub stars, used by 6 labs in the US and Europe.

JAXPythonNeuropixelsBayesian inference

PhD Researcher

Columbia University — Zuckerman Institute

2017 – 2022 New York, NY

Advisor: Prof. Liam Paninski. Scalable inference methods for calcium imaging data in decision-making circuits.

  • Developed a GPU-accelerated spike deconvolution algorithm that is 40x faster than CaImAn on 100K-neuron datasets.
  • First-authored three papers in Neuron, NeurIPS, and PLOS Computational Biology.
  • Thesis: 'Scalable Bayesian methods for neural population decoding' (Zuckerman Prize for Outstanding Thesis, 2022).

PythonPyTorchCUDACalcium imaging

Research Intern

Google DeepMind — Neuroscience Research

2020 – 2020 London, UK

Summer internship applying deep RL models to neural data from the hippocampus.

  • Showed that a successor-representation RL agent predicts hippocampal place-cell remapping patterns with 89% accuracy.
  • Co-authored a PNAS paper with three DeepMind researchers.

PythonTensorFlowReinforcement learning

Selected Publications

  1. Patel, A.; Remington, E.; Jazayeri, M. (2025). Continuous time representation in prefrontal population dynamics. Nature Neuroscience.

    Demonstrates that elapsed time is encoded as a continuous manifold in prefrontal cortex using a novel latent dynamics model.

  2. Patel, A.; Jazayeri, M. (2024). NeuroPLDI: variational inference for neural population latent dynamics. NeurIPS 2024.

    A VAE framework for inferring continuous latent dynamics from large-scale neural recordings. Spotlight paper.

  3. Patel, A.; Paninski, L. (2022). GPU-accelerated spike inference for large-scale calcium imaging. Neuron.

    40x speedup over CaImAn through GPU batching and structured variational approximations.

  4. Patel, A.; Botvinick, M.; Kurth-Nelson, Z.; Behrens, T. (2021). Hippocampal remapping predicted by successor representations in deep RL agents. Proceedings of the National Academy of Sciences.

    Shows that a deep RL agent using successor representations reproduces key features of hippocampal remapping.

  5. Patel, A.; Pillow, J.; Paninski, L. (2020). Scalable Bayesian decoding of neural population activity. PLOS Computational Biology.

    Introduces an amortized variational method for Bayesian neural decoding that scales to 100K+ neurons.

Open-Source Software

Technical Skills

Languages

PythonMATLABJuliaC++RSQL

ML / Stats

JAXPyTorchTensorFlowStanBayesian inferenceVariational methods

Neuroscience

NeuropixelsCalcium imagingSpike sortingPopulation decodingLFP analysis

Infrastructure

SLURMAWSDockerGitJupyterWeights & Biases

Education

Columbia University

New York, NY

2017 – 2022

PhD in Neuroscience (Computational Track)

Coursework & activities

  • Dissertation: 'Scalable Bayesian methods for neural population decoding' (Advisor: Liam Paninski).
  • Zuckerman Prize for Outstanding Thesis, 2022.

Indian Institute of Technology Bombay

Mumbai, India

2013 – 2017

BTech in Electrical Engineering, Minor in Computer Science

  • GPA 9.4 / 10.0
  • Honors Institute Gold Medal. President's Gold Medal for highest GPA in the graduating class.

Languages

  • English Fluent
  • Hindi Native
  • Gujarati Native
  • French Basic

Honours & Fellowships

Teaching

Computational Neuroscience (MIT 9.40)

MIT

Guest Lecturer Spring 2024

Statistical Methods for Neuroscience (Columbia NBHV W4010)

Columbia University

Teaching Assistant (3 terms) 2018–2020

Neuromatch Academy — Computational Neuroscience

Neuromatch

Lead TA Summer 2021