UCLA

Photoplethysmography foundation model whose hierarchical convolutional encoder exposes one embedding per temporal scale and runs entirely on-watch.

Histopathology foundation model aligning H&E image patches with co-registered spatial transcriptomics through a mixture of data experts.

Digital cytology foundation model, a ViT-Base pretrained with iBOT self-supervision on 1.4 million patches from breast, cervical, and thyroid slides.

Histopathology encoder pretrained entirely on prototype-guided synthetic H&E patches, matching models trained on 60-760x more real patient tiles.

Histopathology encoder pretrained on synthetic H&E patches mixed 1:1 with real TCGA tiles, outperforming UNI on lung and lymph node subtyping.

Ternary complex structure predictor for PROTACs and molecular glues, placing E3 ligase, degrader, and target protein in one SE(3)-equivariant pass.

Mixed-modal biomedical assistant that generates medical images as well as text, covering radiology and histopathology visual chat and captioning.

Multimodal RNA chat model pairing a frozen RNA-FM encoder with a Llama-3 8B backbone to answer natural-language questions about a submitted sequence.

Gene expression prediction across a megabase of DNA that aligns frozen regulatory sequence features to language-model tokens by cross-attention.

CLIP-style vision-language model for echocardiography, pretrained on echo videos and cardiologist reports for zero-shot cardiac interpretation.

Drug combination synergy prediction across cancer cell lines, from LLM text embeddings of drugs and cell lines rather than structures or expression.