Our Research
Agentic, Foundational & Spatial AI
Built for Veterinary Medicine
Animal health needs its own AI — not models borrowed from human medicine. Veterinary biology, its radically heterogeneous patients, and the real conditions of practice impose constraints no repurposed model accounts for. We build AI made for veterinary medicine across three modes — foundational, agentic, and spatial — with a focus on pathology and oncology.
Three Pillars
Three AI paradigms organize our work — foundational, agentic, and spatial. Each is built veterinary-first, and together they let us move fluidly between pure software and physically instrumented systems, whichever a problem in pathology and oncology demands.
Foundational AI
Veterinary medicine needs its own foundation models. The field has largely made do with models repurposed from human medicine, importing human-centric assumptions and ignoring what makes veterinary biology distinct. We build veterinary-first models that take the unique biological constraints of animal medicine — and the radically heterogeneous patients a veterinarian cares for, across species, breeds, and body plans — as first-class design problems rather than noise to correct away.
Agentic AI
Agentic systems in veterinary medicine have to operate in the real conditions of practice: distributed across disconnected systems, fed by highly unstructured data, and often without the compute or specialist support taken for granted in human medicine. We build agentic AI that automates clinical workflows and augments clinician capability under exactly these constraints — extending genuine expertise to remote and solo practitioners as specialty access shrinks. Functioning in resource-limited settings is a core design target, not an afterthought.
Spatial AI
A veterinary clinic is a physically heterogeneous environment, and its patients and specimens carry spatial constraints found nowhere in human medicine. Most medical AI flattens this three-dimensional world into text and 2D slices; we build systems that don't. For agentic systems to actually use veterinary foundation models in physical space, they need world models to reason over — reconstructed anatomy and biology with real volume, depth, physics, and geometry, situating AI in the physical structure of veterinary biology.
Building the Future
As a newly established lab, we're assembling the team, partnerships, and infrastructure to build veterinary-first AI for pathology and oncology — and to prove that animal health is better served by models made for it.
Pathology & Oncology Focus
Our pillars converge on pathology and oncology, where representation, autonomy, and spatial reasoning meet the hardest problems in animal cancer care — from tissue and tumor analysis to diagnosis and the digitization of physical specimens.
Built for Veterinary Constraints
We design for the realities of veterinary practice: heterogeneous patients across species and breeds, distributed and unstructured clinical data, and resource-limited settings. These constraints shape our models from the start rather than being patched in later.
Open Data & Reproducible Methods
Veterinary-first AI depends on curated data, interoperable standards, and reproducible pipelines. We build and release these alongside our methods so the broader field can build on them.
Join Our Mission
We're actively seeking graduate students, postdocs, and collaborators excited about building AI made for veterinary medicine. Whether you're coming from computer science, veterinary medicine, biomedical informatics, or pathology — if you're passionate about animal health and real-world impact, we want to hear from you.
Get in Touch