Ovet is an AI-enabled workflow product for veterinary emergency and specialty clinics. I co-founded it with a veterinarian and owned product, design, AI, and the user-facing engineering, while an engineering lead handled the infrastructure and backend. We took it from concept to pilot deployments and first revenue, running thousands of intakes a month.

The problem sits right at the front door of the ER. Triage is slow and manual, the clinical team is stretched thin, and the one person who knows what actually happened, the owner, is usually too stressed to give a clean history. So rather than speed up the same process, we moved the history-taking earlier and let AI turn a worried owner's plain-language story into the structured record a clinician can act on. I designed both halves of that.

For the owner, I designed a guided intake they complete on their phone, in the waiting room while their pet is being seen. It prompts them dynamically, asks the right follow-ups in plain language, and never feels like a form to fill out at the worst moment of their week. The hard part was empathy under stress: keep it calm, short, and human while still pulling the clinically meaningful details out of an anxious, non-expert account.

For the clinic, I designed the other side: a desktop triage dashboard where the team works the live queue by status and elapsed time, and each patient arrives as a structured anamnesis, a reason-for-visit narrative up top with structured clinical findings below, coded against standards like FHIR and SNOMED so it drops straight into their workflow. Underneath both sits the AI pipeline I built, multi-model with a Claude primary and an OpenAI fallback, domain-specific prompt engineering, and structured output schemas with owner validation and priority ranking. It cut a veterinary technician's triage from roughly thirty minutes to about five.