I first attended SAVY 1.0 last April, and was struck by how closely the veterinary AI space mirrors the challenges I’m used to seeing in human health research. Whether it’s claims data, clinical trials, regulatory science, or HTA, the methodological and operational issues are often the same. The experience reshaped how I think about cross-species research and the potential for shared solutions across disciplines.

One Health (the idea that human, animal, and environmental health are deeply interconnected) is a widely recognized public health framework, and it was a clear throughline across the sessions at SAVY 2.0, from oncology to AI ethics to herd health. It’s clear that humans and animals share pathogens, ecosystems, and now, increasingly, data science problems! The most compelling insight for me came from a Google Health plenary on foundation models in medical imaging. One example compared models predicting necrotizing enterocolitis in premature (human) infants: a conventional model scored an AUC of approximately 0.6, while a foundation model (pretrained on adult chest x-rays) reached approximately 0.9. It was a striking example of how large-scale pretraining can lead to impressive performance gains, even when applied to different tissues, ages, or contexts.

Veterinary data is famously small and siloed. In this context, foundation models offer something game-changing: a way to transfer learning from large, unrelated datasets, possibly even from human data, to improve predictions for animal patients. And potentially, vice-versa! Foundation models could be a key to unlocking AI applications in rare conditions, small clinical trials, and early-stage biotech, on both the animal and human side. Translational research came up in different ways across the conference. One example was ImpriMed’s platform for canine lymphoma, which tests live tumor cells from dogs to predict individualized drug responses. Their models are continuously refined using real-world clinical outcomes. It’s a powerful reminder that dogs with spontaneous cancers are not just proxies for humans, they are patients too. The insights gained from treating them have value for both veterinary and human oncology.

At Precision Analytics, this resonates. We spend a lot of time grappling with the same core issues: limited sample sizes, high-dimensional predictors, the need for careful validation, and the practical realities of regulatory science. SAVY 2.0 reinforced that these challenges (and opportunities) span species, and that the future of AI in medicine, human or animal, is collaborative.