Back
Stirrup
Year
2024
Tech & Technique
Flutter (Dart), SvelteKit, TypeScript, Tailwind CSS, Deno, Supabase, Google Gemini AI, Claude AI, PostgreSQL (Supabase), TensorFlow Lite (On-Device ML), Mobile & Web Cross-Platform
Description
An AI-powered fitness training platform combining a native Flutter mobile app, SvelteKit web dashboard, and Deno backend. Features intelligent workout generation, real-time coaching via chat, on-device pose detection with TensorFlow Lite, automatic PR tracking, and a social community feed.
My Role
As co-founder and engineer, I:
- Built a triple-stack architecture: native mobile (Flutter) + web dashboard (SvelteKit) + dedicated API (Deno)
- Integrated Google Gemini and Claude APIs for AI-driven personalised training plan generation
- Implemented on-device pose detection with TensorFlow Lite and OpenCV for real-time form analysis
- Designed offline-first sync with local Isar database and automatic cloud reconciliation
- Built a triple-stack architecture: native mobile (Flutter) + web dashboard (SvelteKit) + dedicated API (Deno)
- Integrated Google Gemini and Claude APIs for AI-driven personalised training plan generation
- Implemented on-device pose detection with TensorFlow Lite and OpenCV for real-time form analysis
- Designed offline-first sync with local Isar database and automatic cloud reconciliation