Join Us

The Fries Lab works on foundation models for healthcare, focusing on multimodal data (structured EHRs, clinical text, imaging, genomics), longitudinal reasoning over patient trajectories, benchmarking, and human-AI teaming. A central emphasis of the lab is data-centric AI. Model capabilities are shaped by the data used to train them, motivating careful study of how bias, noise, data mixtures, and supervision affect model behavior. Data quality often matters more than data quantity! We study how to train, evaluate, and deploy these models in realistic clinical settings. This includes developing new training objectives, building benchmarks that reveal where and why models fail, and designing human-AI feedback loops that integrate expert judgment and clinical workflows. Our goal is to make clinical context explicit and measurable, particularly under noise, sparsity, and domain shift.

Our work has appeared in NeurIPS, ICLR, AAAI, npj Digital Medicine, Nature Communications, and Nature Medicine. We work closely with hospital data science and clinical partners—offering postdocs a rare opportunity to do foundational AI research while engaging directly with healthcare systems.