Projects
What I built, and why it matters, in plain language.
I build theory-driven, neuroscience-inspired models of neural circuits, and the deep learning tools to analyze them. My first-author eLife paper derives a hierarchical recurrent circuit theory that unifies divisive normalization, gamma-band coherence and inter-areal communication subspaces.
I work hands-on with stochastic dynamical systems, recurrent neural networks, spectral and information-theoretic analysis of time series, and generative models (autoencoders, VAEs, GANs), implemented end to end in Python and PyTorch. I am a Ph.D. candidate at NYU, advised by Prof. Stefano Martiniani and collaborating with Prof. David J. Heeger.
| Modeling & theory | Recurrent neural networks, stochastic differential equations, dynamical systems & stability/bifurcation analysis, mean-field and neural-field models, divisive normalization (ORGaNICs) |
| Signal & data analysis | Power spectra & coherence, communication subspaces / dimensionality reduction, mutual information & KL-divergence methods, time-series analysis, transcriptomic data analysis |
| Machine learning | PyTorch, deep generative models (autoencoders, time-lagged AEs, VAEs, GANs), scikit-learn, end-to-end training & evaluation pipelines |
| Programming & tools | Python (NumPy, Pandas, Matplotlib), SQL, Fortran, shell; MATLAB, Mathematica; Git, VS Code, Cursor |
A. Pal, S. Rawat, D. J. Heeger, S. Martiniani. Hierarchical neural circuit theory of normalization and inter-areal communication. eLife (Reviewed Preprint), 2026.
In plain words: a mathematical model of how connected brain areas talk to each other. One circuit, with one set of parameters, explains both the brain rhythms and the low-dimensional activity patterns that experiments had linked to communication between areas, and predicts how top-down feedback strengthens that communication.
J. Han, A. Grau-Perales, R. M. Harris, E. Lesburguères, H.-Y. Kao, A. Pal, J. M. Alarcon, T. C. Sacktor, S. Martiniani, H. A. Hofmann, A. A. Fenton. Persistently increased expression of PKMζ and unbiased gene expression profiles identify hippocampal molecular traces of a long-term active place avoidance memory and “shadow” proteins. Advanced Science, 2026.
In plain words: some molecules that are essential for keeping a memory do not show up when you simply compare gene expression levels. Our method, C-SCoRe, looks at how relationships between genes change instead, and recovers the link between those genes and how well mice remember.
A. Pal, S. Pal, S. Verma, M. Shiga, N. N. Nair. Mean force based temperature accelerated sliced sampling: efficient reconstruction of high dimensional free energy landscapes. J. Comput. Chem. 42, 2021.
In plain words: a faster way to turn molecular simulations into a map of which molecular shapes are stable, which removes a slow and costly reweighting step.
Ph.D. in Chemistry (research: computational neuroscience). Advisor: Prof. Stefano Martiniani; collaborator Prof. David J. Heeger. Henry M. MacCracken Doctoral Fellow.
M.S. in Chemistry.
B.S. in Chemistry.
What I built, and why it matters, in plain language.
Papers, posters and talks with links.
Including an interactive explainer of the normalization circuit.
Download my CV as a PDF.