Wellhound
At-Home Preventive Dog Health (UVA Darden i.Lab)
Overview
Wellhound brings preventive blood work to a dog's home. A licensed technician draws blood at the owner's door, a partner reference lab runs the panel, a veterinarian reviews the results, and the owner gets a care plan plus an AI assistant that knows their dog's history.
I'm the founder and product lead. I wrote the PRD, ERD, and data flows, designed the database, and built most of the app myself across 100+ pull requests.
Problem
Dogs hide illness. Owners read early signs as aging, and most disease gets caught late, at the point where it is expensive or untreatable. Annual vet visits are a snapshot, and nobody wants to go to the vet more often.
What owners told me they want: keeping their dog healthy for longer, at home, at a price that makes sense, with a verified opinion and no extra hassle.
Discovery & Pivot
I interviewed 30 dog owners (senior-dog owners, at-risk breeds, new owners, and owners who had lost a dog to something found late) and 10 veterinarians.
The first version was a quarterly mail-in urine and fecal kit. Vets were consistent: blood is where the diagnostic value is. That feedback drove the pivot from mail-in kits to in-home blood testing, with testing panels designed alongside a practicing veterinarian.
Product
• Signup, membership payments, and at-home blood draw booking
• Results stored against each dog's own baseline, with plain-language explanations of every biomarker
• Vet-reviewed care plans (AI drafts, the vet has the final say)
• An AI care assistant grounded in the dog's record
• Owners can invite their own vet to receive results and updates
AI Systems
• Model routing: Claude Haiku handles routine queries and Sonnet handles complex, tool-using workflows. About 75% of requests route to Haiku, cutting estimated inference cost by about 50%
• Guardrails: the assistant is scoped away from giving veterinary advice and routes clinical questions to a vet
• Synthetic customer testing: AI testing agents grounded in real dog-owner reviews found 96 pre-launch defects and raised simulated checkout completion by 20 percentage points
• A reference-linked knowledge base that consolidates Slack, Notion, advisor feedback, and product docs into structured Markdown context for AI-assisted development
Stack
Next.js (App Router, TypeScript), Supabase, Stripe, Resend, PostHog, Vercel, Claude API.
Status
Won a $5K UVA Darden i.Lab grant. Recruited veterinary advisors, built a reference lab partnership, and led a 4-person team preparing a paid November 2026 beta in Charlottesville, Virginia.
What I'm Learning
• Regulation shapes the product: who can draw blood, who can interpret it, and how a vet relationship is established decide the architecture before any feature does
• Operations should be automated unless they literally can't be
• The hardest open question is willingness to pay, which is exactly what the beta is built to test