DogBase

I’m a K9 search and rescue handler. My dog is a Belgian Malinois named Echo. In the field I kept hitting the same wall: the records that prove a dog is ready for deployment (training logs, deployments, health, certification) were scattered across tools that were never built for working dogs. DogBase started as the tool I needed myself.

The buyers are law-enforcement, government and SAR organisations. Handlers work outdoors with patchy connectivity, and procurement asks hard questions: GDPR, role-based access, encryption at rest and in transit, audit logging. The product had to work offline in the field and pass a security review in the office.

Run lean, own the stack, and put AI where it saves a handler real time.

• →Offline-first Flutter mobile app with BLE sensor integration, because the field has no signal.
• →Go microservices on Cloud Run with event-driven Firebase and BigQuery data infrastructure, so a small team can run it.
• →AI shipped as product, not as a demo: Intelligent Journal (LLM via Vertex AI and Gemini), AI Vision (computer vision on dog-behaviour video), and a RAG architecture for the AI co-pilot.
• →Replaced most paid SaaS with self-hosted open source, which cut operating costs by about 75%.

The platform end to end: the offline-first Flutter app (iOS and Android), the SvelteKit web app, Go microservices and the data layer, plus the AI features. I also led technical discovery, demos, solution design and pricing for buyers in multiple countries, and built and led a distributed team of 10 engineers.

Working-dog teams in 10+ countries use DogBase, with 350K+ training minutes logged (up 39% in six months), 35% signup-to-active conversion and ~70-minute average sessions, all with zero paid acquisition. It went through Techstars in 2023, won the Latitude59 2026 pitch competition (465 startups from 53 countries), and is part of NVIDIA Inception. I’ve presented it at the World Police Summit, Frontex and the International Police K9 Conference.