Things I'm building

Mosaic - Inferring multi-scale molecular features from digital pathology slides. Funded by the ICI and aiTDIF. Featured at AACR AI and ESMO AI. Check out the code here.
Orpheus - Multimodal AI using clinical text and pathology slides to identify high-risk breast cancer and generate synthetic breast cancer images conditioned on risk. The code is here.
Mussel - Highly scalable preprocessing for digital pathology, built with Raymond Lim.

About me

I’m a radiation oncology resident at Memorial Sloan Kettering and a postdoc in Computational Oncology, where I work with Nikolaus Schultz, Francisco Sanchez-Vega, and Sohrab Shah and lead the Pathology Data Mining team.

There’s a ton of data generated during routine cancer care — pathology slides, sequencing panels, clinical records. I build AI to find useful patterns in these data. I’m especially interested in multimodal modeling, where there’s a massive opportunity to connect what we see under the microscope with what’s happening at the molecular level and how, taken together, that affects response to therapy.

Perspectives

Harnessing multimodal data integration to advance precision oncology - Roadmap to develop multimodal AI for oncology.
Simplifying clinical use of TCGA molecular subtypes through machine learning models - Consistent tumor subtype identification with parsimonious genomic data.

Previous work

Late-fusion multimodal AI to risk stratify ovarian cancer - Combining histopathology, radiology, and clinicogenomics for ovarian cancer prognosis.
Predicting peptide presentation by major histocompatibility complex class I - Improved deep learning for MHC class I peptide presentation.
Resonance - Annual science conference connecting Yale researchers with New Haven high schoolers.

Affiliations

Current
Previous

Interests

Metal sculpture / Brooklyn Makerspace Techno Green tea Old-school bodybuilding gyms Gaming Hiking Fantasy novels

Languages

English (Native) Spanish (Professional) Japanese (Intermediate) Swedish (Beginner)