Subhrajit Dey

hover over me @ nyu.edu

New York University

370 Jay Street, 11th Floor
Brooklyn, NY

Biography

I am a full-time Research Engineer in the NYU Neuroinformatics Lab, working with Prof. Erdem Varol. I recently graduated with an M.S. in Computer Science (Neuroinformatics) from NYU Tandon School of Engineering. I build tools and analyze large-scale neural data spanning brain–computer interfaces (BCIs), Neuropixels electrophysiology, and related multimodal recordings, with collaborators including Prof. Saurabh Vyas and Prof. Attila Losonczy. My research interests center on foundational models for neuroscience and multimodal neural data translation.

Longer term, I want to help build brain implants and neural interfaces that let people control and communicate with AI agents and the Internet as fluidly as imagined in Neuromancer and Cyberpunk 2077—technologies in the spirit of neural lace, direct brain–computer links, and closed-loop neuroprosthetics.

Before joining the lab full-time, I completed my M.S. at NYU Tandon and worked as a Computer Vision Engineer at Bigvision.ai, building end-to-end computer vision systems for retail and e-commerce. I earned a B.E. (Honours) in Electrical Engineering from Jadavpur University, where I published work in medical imaging and computer vision. During my master’s, I also served as a Course Assistant at NYU Stern (Consulting Capstone, Corporate Finance, and Foundations of Corporate Finance, Summer 2025), and developed a CRE loan-financing workflow with multi-LLM agents under Prof. Anthony Marciano.

Research

My research interests are in neuroinformatics: foundational models for neural data, multimodal translation across electrophysiology, imaging, and transcriptomics, representational alignment between the brain and large world models, and tooling for BCIs and Neuropixels-scale recordings.

I currently work on self-supervised pretraining and post-training of foundation models for Neuropixels LFP (geometry-aware SSL, masked prediction, and localization heads for 3D atlas coordinates and brain-region decoding), and on cross-modal cell registration linking in-vivo two-photon imaging, ex-vivo microscopy, and spatial transcriptomics. Some recent representative papers are below; others are on my Google Scholar profile.


Research Engineer NYU Neuroinformatics Lab 2025 - Present	Computer Vision Engineer Bigvision.ai 2022 - 2024	Mitacs Research Intern ÉTS Montreal Summer 2021	B.E. Electrical Engg. Jadavpur University 2018 - 2022

Updates

[2026] Graduated with an M.S. in Computer Science (Neuroinformatics) from NYU Tandon; now a full-time Research Engineer in the NYU Neuroinformatics Lab.
[2026] Submitted "Geometry-Aware Self-Supervised Electrophysiology Representation Learning" to NeurIPS 2026 (under review).
[Apr 2026] Preprint out — "A transcriptomic axis aligns with in vivo functional dynamics in hippocampal inhibitory circuits" (Yong, Herrlinger, …, Dey, Varol, Losonczy); submitted to Nature/Science. Combines 2-photon calcium imaging & spatial transcriptomics in the hippocampus of behaving mice, with a 4-modality 3D viewer. [bioRxiv] [Website]
[May 2025] Started my second year of MS in Computer Science at NYU Tandon with a perfect 4.0 GPA! 🔥
[Spring 2025] Started working as a Research Assistant in NYU Neuroinformatics Lab working on multimodal neural data translation models with Prof. Erdem Varol.
[Spring 2025] Served as a Course Assistant for Consulting Capstone, Foundations of Corporate Finance, and Corporate Finance at NYU Stern.
[Sep 2024] Started my Master's journey at NYU Tandon School of Engineering! Excited to dive deep into AI research.
[June 2024] Completed 2 years as Computer Vision Engineer at Bigvision.ai
[Dec 2022] Published "Cross-Lingual Sentiment Analysis using Transliteration and Character Embedding" at ICON-2022, ACL! [Paper]
[Jan 2022] Published "Optimized Fuzzy Ensemble of CNNs for TB Detection" in Applied Soft Computing, Elsevier! [Paper] - 45 citations and counting!
[Sep 2021] Completed Mitacs Globalink Research Internship at ÉTS Montreal, developing 360StereoNet for depth estimation.
[June 2022] Graduated from Jadavpur University with B.E. (Hons.) in Electrical Engineering with 8.78/10.0 GPA.

Publications

2026

Figure 1: Geometry-aware SSL pipeline overview

Geometry-Aware Self-Supervised Electrophysiology Representation Learning

S Dey, P-J Lu, V S Parupudi, M Patel, M-E Conde Paredes, M Vöröslakos, A Maslarova, G Buzsáki, S Vyas, E Varol

NeurIPS 2026 (Under Review)

Self-supervised framework that learns geometry-aware representations of single-channel LFP by using probe channel geometry as a supervisory signal. Compares speech SSL objectives (e.g., masked prediction) with geometry-aware losses that mirror physical channel layout; evaluates Wav2Vec 2.0, Whisper, and Data2Vec against supervised baselines across datasets, labs, species, and probe technologies for 3D coordinate regression and brain-region classification.

PDF Under Review

A transcriptomic axis aligns with in vivo functional dynamics in hippocampal inhibitory circuits

H C Yong, S A Herrlinger, M E Conde Paredes, C K O'Toole, J Yoo, B Y Rao, T S Mihaila, J Shi, S Dey, E Varol, A Losonczy

bioRxiv preprint; Submitted to Nature/Science, 2026

End-to-end pipeline linking cell-resolved two-photon calcium imaging in behaving mice to spatial transcriptomics in hippocampal CA1 interneurons, with a 4-modality 3D viewer for cross-modal registration.

bioRxiv Website