About Me

I am a Junior Fellow of the Harvard Society of Fellows, an affiliate of the Harvard Center for Brain Science, and a recipient of a 2024 NIH Director’s Early Independence Award. My research interests lie in some neighborhood of the intersection between theoretical neuroscience, statistical physics, and machine learning. For more information about me and my work, see some of my recent publications and lecture notes, or my CV.

I will be joining The Rockefeller University as an Assistant Professor and Head of Laboratory in January 2027. Multiple postdoctoral positions in my group will be available; please get in touch if interested and stay tuned for a formal advertisement.

News from the past year

• [May 2026] Our paper with Alex Atanasov, Blake Bordelon, Courtney Paquette, and Cengiz Pehlevan on high-dimensional asymptotics of stochastic gradient descent in linear random feature models has just been published in ATMP.

• [May 2026] Our synthetic paper with Alex Atanasov and Cengiz Pehlevan on high-dimensional ridge regression has just been published in JSTAT.

• [March 2026] Juan Carlos’ paper on efficient coding in early olfaction (with Venki Murthy and Farhad Pashakhanloo) has just been published in PNAS! See also the Harvard MCB and Kempner Institute news articles.

• [March 2026] If you’re at COSYNE in Lisbon, check out Juan Carlos’ poster 1-007, based on our paper on efficient coding in early olfaction.

• [January 2026] Our note on disordered kinetic Ising models as a toy model for sequence data with tunable correlations has now been published in JPA.

• [December 2025] I’m headed to NeurIPS! If you’re in San Diego, check out our UniReps workshop paper with Farhad Pashakhanloo on how data symmetries affect representational similarity analysis, our UniReps perspective based on our long-form piece on summary statistics of learning, and our NeurReps proceedings paper on how training shapes the Riemannian geometry of deep network representations.

• [September 2025] I’m headed to the ROC(K)IN’ AI 2.0 workshop, where I’ll speak about our work on generalized cross-validation in high-dimensional ridge regression with correlated samples.

• [August 2025] Our perspective on summary statistics of learning—and what lessons the statistical physics of learning might have for the analysis of neural data—has just been published.

• [August 2025] I’m speaking at the 2025 Brains, Minds, and Machines Summer Course at the Woods Hole MBL.

See older news here.

I can also occasionally be found on Bluesky or (very rarely) on Mastodon. In official contexts I use my full name, so you’ll find me listed as Jacob Andreas Zavatone-Veth.