TailZaar AI Pet Assistant on Nemotron

Inspiration

We're pet care first, an AI pet care companion app plus a pet marketplace. We do have a marketplace option, so I think we have the mix of both.

One can have multiple pets. They can log their pets in full. They can see what's due. The health hub has vaccinations and ongoing medications, and you can log these as well, and they can remind you and can send you notifications and reminders. Also, there is like a pet crib where you can see your pet within its home. PetCare is fully there, an AI assistant. Everything is working. The marketplace is currently complementary to these features.

When this hackathon came up, our CEO's word to us was one line: "we can use any NVIDIA model and put it in the app, we don't have to change actually anything."

What it does

TailZaar is a pet care app for pet owners anywhere, with a marketplace for Riyadh, on the App Store since 2 October 2026 and at version 1.6.4 today, in Arabic and English. I am its co-founder and CEO.

Pet care comes first. An owner keeps one or more pets in the app, each with a full profile and a journal. The health hub holds vaccinations and ongoing medications and sends reminders when something is due. There are lessons and daily tips, a symptom checker, and a pet crib where you see your pet in its home. The AI pet assistant sits in the middle of all this: an owner asks it a pet care question in a chat, and it answers from that pet's own record. Behind the pet care there is a marketplace, with products, services with online video consultations, and bookings in Riyadh.

For this hackathon we built one specific thing for NVIDIA and Nebius: a second model route for the assistant. In the backend service that serves the assistant, a feature flag now selects NVIDIA Nemotron served on Nebius Token Factory as the model that writes the assistant's answers. The flag is off by default. The route is backend only, and the app on the App Store did not change. Token Factory gives us a hosted Nemotron endpoint behind one flag, so the model behind the assistant can change without the app changing. The thin wrapper service that makes the Token Factory call is public under the MIT licence; the rest of TailZaar stays closed.

I switch the Nemotron route on for the judging period, 1 to 15 December 2026. Judges need nothing from us: install TailZaar from the App Store, sign up with your own phone number, open the AI pet assistant and ask it a pet care question. During the judging period the answer comes through Nemotron on Token Factory.

How we built it

Customer app: React Native and TypeScript, on iOS. Backend: NestJS, TypeORM and PostgreSQL, one backend for every surface, the customer app, the partner app, the dashboards and the web storefront at tailzaar.com.

The hackathon change lives in the backend only. A feature flag selects the Nemotron route, and it is off by default. In our CEO's words: "The IP is ours. We can open source the repo part. But we won't open source the rest of the app." So the public repository holds only the Nemotron-backed assistant service wrapper: the call path, its configuration, a README and the MIT licence file; the TailZaar product code is private and stays out of the repository.

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for TailZaar AI Pet Assistant on Nemotron

The Nemotron route is built as a flagged serving option for the assistant, switched on per environment and per account; it stays in the assistant's backend service behind its flag, in the environments where it is deployed.

Built With

• ios
• nebius-token-factory
• nestjs
• nvidia-nemotron
• postgresql
• react-native
• typeorm
• typescript