I’m a PhD student in computer science at UCLA, advised by Sriram Sankararaman. I build foundation models and statistical methods for biobank-scale health data. I use both to understand the biology behind disease.
My work is supported by an NIH T32 fellowship and an Anthropic AI for Science grant. Outside research, I paint (a few pieces are here), play piano and French horn, and dance hip-hop.
Research
Biobanks now pair genomes with many kinds of health data for hundreds of thousands of people, including wearable sensors, medical imaging, lung function tests, blood biomarkers, and health records. I build foundation models that learn from these data and statistical genetics methods that find how genes shape health and disease.
My recent modeling work is on wearable sensors. I co-led Inertia-1 (NeurIPS 2026) with Yuzhe Yang’s lab, an open study of how to pretrain motion foundation models on 18 million hours of accelerometry. Our follow-up study tests the model on diagnosis, prognosis, and disease progression, and measures how much genetic signal its learned representations share with disease.
To tie modeling and genetics together more directly, I’m developing a training loss that pushes a model’s learned traits to be heritable and genetically correlated with disease, so a genetic study of those traits can find loci that standard measurements miss. It’s designed to work with any model and any data type, from imaging to spirometry.
On the genetics side, I’ve helped develop fast variance-component methods for gene–gene interactions (FAME, QuadKAST), gene–environment interactions (ENGINE), and rare-variant heritability (FLEX). Current projects extend them to rare disease, single-cell data, and genetic correlation across tissues.
News
- Inertia-1, accepted to the NeurIPS 2026 main track, will also be an oral at the GenAI for Health workshop at NeurIPS 2026 in Sydney. I’m presenting it in December.
- Inertia-1 was accepted to NeurIPS 2026.
- Started an NIH T32 fellowship with UCLA’s Genomic Analysis Training Program.
- ENGINE is out in Genome Research, selected from RECOMB 2026.
- Received an Anthropic AI for Science grant for rare disease research.
- Posted Inertia-1 on arXiv, with open code.
- Finished my M.S. in Computer Science.
- Metapipeline-DNA is out in Cell Reports Methods.
- FAME is out in Nature Genetics. UCLA Health wrote about it.
- Posted FLEX on bioRxiv.
Selected publications
- Inertia-1: An Open Exploration of Wearable Motion Foundation Models NeurIPS 2026 main trackalso an oral at the GenAI for Health workshopcodeproject
- A biobank-scale method for learning modulators of gene-environment interaction underlying human complex traits from multiple environmental exposures Genome Research 2026selected from RECOMB 2026code
- A biobank-scale test of marginal epistasis reveals genome-wide signals of polygenic interaction effects Nature Genetics 2025codepress
- A scalable adaptive quadratic kernel method for interpretable epistasis analysis in complex traits Genome Research 2024selected from RECOMB 2024code