News: NVIDIA Academic Grant Awarded for SAFE-Vet

News | NVIDIA Academic Grant Program

SAFE-Vet—Signalment-Conditioned Alignment and Clinical Safety Benchmarking for Veterinary Edge AI—has been selected for the NVIDIA Academic Grant Program. In support of this work, NVIDIA is donating 1x DGX Spark to the University of Wisconsin–Madison to accelerate research on clinically grounded, safety-focused generative AI for veterinary medicine.

SAFE-Vet addresses a central challenge in veterinary GenAI: safety-critical reasoning often depends on structured patient context such as species, breed, and age. Our work studies whether explicitly conditioning small, edge-deployable language models on veterinary signalment can improve triage accuracy, reduce harmful hallucinations, and avoid failure modes such as overly defensive refusal behavior. The project introduces VetSafetyEval, a benchmarking framework built around high-stakes veterinary triage scenarios, and evaluates how different conditioning strategies affect accuracy, safety recall, and clinical utility.

This award will support continued development of safe, transparent, and deployable veterinary AI systems—particularly for settings where reliable cloud connectivity is limited and lightweight local models may offer practical advantages for clinical access and equity.

NVIDIA’s support will help expand benchmarking, model development, and hardware-enabled experimentation for the next phase of SAFE-Vet.

Related Update | AMIA 2026 Amplify Informatics Conference

SAFE-Vet was also accepted for poster presentation at the AMIA 2026 Amplify Informatics Conference.

• Presentation Date: May 20, 2026
• Time: 5:00 PM–6:30 PM MDT
• Session: Poster Session 2 and Reception
• Location: Grand Hyatt Denver

We are grateful for this support and recognition as we continue building clinically useful and safety-conscious AI for veterinary medicine.