Cole Hurwitz

We are hiring on the IBM Core AI team. Feel free to DM me. 🙂 Job posting form - forms.gle/zZ2FHDg5sPVq...

Our group builds the technology that makes agentic AI reliable, observable, and governable at enterprise scale, with open research and tooling for evaluating agents and improving them.

After nearly a decade in academia, I am thrilled to share my next chapter: I am joining IBM as an AI Architect in the new Core AI group.

We are building an AgentOps platform to observe, evaluate, and optimize enterprise AI agents and we are hiring. DM me if interested.

Very happy for this to finally be published! We developed new machine learning methods for scalable mapping of synaptic connectivity using holographic optogenetics and compressed sensing.

www.nature.com/articles/s41...

We present our preprint on ViV1T, a transformer for dynamic mouse V1 response prediction. We reveal novel response properties and confirm them in vivo.

With @wulfdewolf.bsky.social, Danai Katsanevaki, @arnoonken.bsky.social, @rochefortlab.bsky.social.

Paper and code at the end of the thread!

🧵1/7

Two flagship papers from the International Brain Laboratory, now out in ‪@Nature.com‬:
🧠 Brain-wide map of neural activity during complex behaviour: doi.org/10.1038/s41586-025-09235-0
🧠 Brain-wide representations of prior information in mouse decision-making: doi.org/10.1038/s41586-025-09226-1 +

Excited to co-organize our NeurIPS 2025 workshop on Foundation Models for the Brain and Body!
We welcome work across ML, neuroscience, and biosignals — from new approaches to large-scale models. Submit your paper or demo! 🧠 🧪 🦾

Excited to announce the Foundation Models for the Brain and Body workshop at #NeurIPS2025! 🧠📈 🧪

We invite short papers or interactive demos on AI for neural, physiological or behavioral data.

Submit by Aug 22 👉 brainbodyfm-workshop.github.io

Neural Encoding and Decoding at Scale (NEDS) is now accepted to @icmlconf.bsky.social as a spotlight (top 2.6%)! 🧠 🧪

Super cool. 😀 Exciting to see the practical use cases of electrical stimulation for treating neurological disorders

We’re hiring postdocs to join my lab at UNC. If you’re interested in adolescence, brain, and social development, DM me. Our work incorporates fMRI, social media, and longitudinal methods. We study risks and opportunities in adolescence. If you’re at #SRCD2025 and want to meet, please reach out!

I asked "on the other platform" what were the most important improvements to the original 2017 transformer.

That was quite popular and here is a synthesis of the responses:

Whoever is at the @iclr-conf.bsky.social workshops, feel free to reach out to meet! Looking for fun neuro conversations since there aren’t any neuro workshops 😢

Scaling models across multiple animals was a major step toward building neuro-foundation models; the next frontier is enabling multi-task decoding to expand the scope of training data we can leverage.

Excited to share our #ICLR2025 Spotlight paper introducing POYO+ 🧠

poyo-plus.github.io

🧵

It’s been great working with you 😄

@steinmetzneuro.bsky.social

Eva Dyer, Chandramouli Chandrasekaran, Nicholas A. Steinmetz, and Liam Paninski (ran out of characters!)

This work was led by @hanyu42.bsky.social who tirelessly worked to make this possible. In collaboration with Hanrui Lyu, Ethan Yixun Xu, @mostsquares.bsky.social, @kenjilee.bsky.social, Fan Yang, Andrew M. Shelton, Shawn Olsen, Sahar Minavi, Olivier Winter, @intlbrainlab.bsky.social, and

We are still working on the codebase and aim to release a tool soon that users can download, fine-tune, and apply to their own datasets!

We evaluate NEMO on brain region localization by predicting the region of individual neurons (and nearby groups) using only the extracted features, and compare it to baseline methods.

NEMO again outperforms both the VAE-based and supervised approaches.

We scale NEMO to the full IBL Brain-Wide Map dataset: 675 insertions from over 100 animals, yielding 37,017 high-quality neurons.

Without using any labels, NEMO's features align closely with anatomical regions and are consistent across labs.

We benchmark NEMO against two SOTA cell-type classification methods, PhysMAP and a VAE (Beau et al., 2025), using two optotagged datasets from the mouse cerebellum and visual cortex.

NEMO outperforms all baselines, including fully supervised models, with minimal fine-tuning.

We construct a paired dataset of spike trains and waveforms for all neurons, transforming spiking activity into an ACG image (Beau et al., 2025) that captures autocorrelation across firing rates.

NEMO is trained to align ACGs and waveforms in a shared embedding space.

In vivo cell-type and brain region classification via multimodal... Current electrophysiological approaches can track the activity of many neurons, yet it is usually unknown which cell-types or brain areas are being recorded without further molecular or...

Building on current multimodal cell-type classification methods (Lee et al. 2024 and Beau et al. 2025), we introduce a contrastive learning method for spiking activity and extracellular waveforms called NEMO. 🐟

Paper: Paper: openreview.net/forum?id=10J...
Website: ibl-nemo.github.io

Thrilled to share our state-of-the-art method for in vivo cell-type classification and brain region localization, NEMO, which is now now a spotlight at @iclr-conf.bsky.social !

We use NEMO to characterize the electrophysiological diversity of cell-types across the entire mouse brain. 🐭 🧪 🧠

Agreed! But here's a note of caution: in the brain, different behavioral contexts can engage completely different neurons! Julie Lee in our lab published this in 2022 (and I'm still digesting the implications).
"Task specificity in mouse parietal cortex"
www.cell.com/neuron/fullt...

Wow, this is fascinating. Thanks for sharing!

This point is really important. We will need to train models across diverse behavioral contexts, rather than over-interpreting results from a single experimental setup or task!

Totally agree. Data is the bottleneck in neuroscience, and far more costly than compute.

But if we find that performance saturates with more data, it might reflect under-parameterized models. So both are key to interpret scaling.

@mehdiazabou.bsky.social and I have discussed this a lot and he has given me a lot of good feedback on this topic.

The Chinchilla paper (Training Compute-Optimal Large Language Models, Hoffmann et al., 2022) is widely regarded as a gold standard for empirically characterizing and optimizing scaling laws for large language models - arxiv.org/pdf/2203.15556