Jordan M. Sorokin

Research

Publications, preprints, and selected highlights

A selection of my peer-reviewed publications, preprints, and conference work — spanning circuit neuroscience and epilepsy, respiratory and arousal circuits, and machine-learning models for biology.

Neurophysiology & Epilepsy Research

My scientific foundation was built at Stanford, where I completed my PhD in Neuroscience (2019) in the Huguenard Lab. My doctoral research focused on the neural dynamics underlying absence epilepsy — specifically, how the thalamocortical circuit transitions into and out of seizure states.

Using a combination of large-scale extracellular recordings, optogenetics, and applied mathematics, I demonstrated that real-time switching of thalamic firing modes is sufficient to both initiate and abort generalized seizures in freely behaving rodents. My work identified pre-ictal oscillatory signatures — including β-band shifts in thalamocortical synchrony — that reliably predict seizure onset, and showed that thalamic activity patterns unfolding across multiple time scales encode the imminence of ictal transitions.

This work produced several high-impact publications, including a first-author study in Neuron (2017) demonstrating bidirectional optogenetic control of generalized epilepsy networks, and a preprint on multi-scale thalamic dynamics as predictors of absence seizure onset.

Biotech, Deep Learning & Current Work

Over the past seven years, I've worked at the intersection of machine learning and biology in industry. At Herophilus and subsequently at Recursion, I developed and deployed deep learning representation models for biological state characterization — working across organoid imaging, transcriptomics, and high-content phenotypic screening.

My work at Recursion spanned self-supervised transcriptomics foundation models (contributing to the TxFM architecture, presented at ICLR 2026 Workshop FM4Science), large-scale representation learning for cancer and neurological disease, and the design of scalable ML infrastructure for drug discovery programs. I led disease program teams and guided cross-disciplinary scientific strategy at the boundary of computational biology and translational research.

I'm now a researcher at Amae Health, where my focus shifts back toward the brain — this time through the lens of passive digital biomarkers. My work there centers on machine learning approaches to inferring latent mental health states from observed behavioral signals, including passive biometrics and speech-to-text models. The goal is to enable richer, more continuous measurement of mental health outside the clinic.

Selected highlights