FIG. 1 — 1° → 3°, folding

1°

Jonathan Zhao

I'm a student at Stanford studying Math & CS.

My current research spans virtual cells and explicit regularisation. And I'm broadly interested in AI for science and ways to understand broad ranges of data.

Previously, I've done research in protein language modelling @Pumpkinseed, paralog dependency @Stanford and water pollution with the UK government, where I won the Stockholm Junior Water Prize and was featured in Forbes.

In my free time, I sing Baritone for Stanford O-tone and compete for Stanford (and prev. England) Debate.

NOW
Undergraduate, Stanford
CURRENTLY
Arc Institute
INCOMING
Bridgewater Associates
IN ANOTHER LIFE
Organist at Duomo Cathedral
LISTENING TO
The New Abnormal
2°

Work

4 roles
2b

Arc Institute

ML Research Intern · part-time

De novo protein design.

Sep 2026 – presentongoing
2c

Stanford Cancer Institute

Computational Biology Researcher

Virtual cells, cancer aneuploidy and ML for cancer vulnerabilities. PIs: Prof Jason Sheltzer & Dr Yi Bei.

Oct 2025 – presentongoing
2d

Pumpkinseed

Machine Learning Engineer Intern

Protein sequence recovery from noisy signal data to 96.7% accuracy: dilated residual CNNs with BERT-style masked pretraining, a CRF output layer with Viterbi decoding, and ESM-2 embeddings. Models proposed to DARPA.

Jun – Aug 2026completed
3°

Research

3 projects
3a2025 – PRESENT

Cancer Aneuploidy

Stanford Cancer Institute · Sheltzer Lab

Virtual cells and ML for cancer vulnerabilities: single-cell virtual knockouts on paralog dependency, across 11 datasets (2.6M cells) on Stanford HPC. PIs: Prof Jason Sheltzer & Dr Yi Bei.

Virtual Cell ↗Expression ↗RNA/DNA ↗
3b2024 – 2025

PFAS Pollution

CIWEM / Xylem / SIWI · preprint

KNN-RF ensemble predicting PFAS concentrations to within 94% of field-measured values across river sources, translated into a point-of-use filter design validated against USEPA software.

Stockholm Junior Water Prize

Forbes ↗SIWI ↗Wiki ↗
3c2023 – 2024

Alzheimer's Pathology

Cambridge Centre for Misfolding Diseases · Cambridge CMD

Large-scale molecular dynamics of protein–ligand interactions and a docking-based virtual screen for β-secretase inhibitors in Alzheimer's.

archived
4°

Fragments

all 7 entries →

I write about neuroscience, philosophy, and occasionally poems of love.

Genetic Anticipation of Huntington's DiseaseWhy does the HTT CAG repeat expand across generations, and why almost always through the father? The case for a repair-driven, not replication-driven, mechanism.No. 07
Parkinson's Disease Is to Lewy Body Disease as Alzheimer's Disease Is to Frontotemporal DementiaThe analogy is tempting — a signature protein anchoring a whole disease family in both cases — but only one side of it is actually a subset relationship.No. 06
Parkinson's and the Chemicals of the WestParkinson's tracks industrialization too closely to be coincidence. Pesticides, solvents, and air pollution all converge on the same loaded gun: the dopamine neuron.No. 05
The Curious Case of ARPdVA Colombian woman carried a mutation that should have caused Alzheimer's in her 40s and a genetic quirk that kept her lucid into her 70s — what her brain reveals about where amyloid stops and neurodegeneration begins.No. 04
Biomarkers of Alzheimer's DiseaseA survey of amyloid, tau, CSF, PET, and blood-based biomarkers for Alzheimer's — and why they're better at ruling pathology in than at predicting exactly when it will strike.No. 03
Origins of Apoptosis: Mitigating Injury or Regulating Development?Apoptosis prunes overproduced neurons in development and quietly disposes of injured ones — but which job did the machinery evolve to do first?No. 02
Hypoxia-Ischemia, Seizures, and HypoglycemiaThree very different neurological insults converge on the same excitotoxic pathway to necrotic neuron death, but they don't damage the same parts of the brain.No. 01