Run App — Fitness Platform
My first project at DRCODE (as an intern), built ground-up: a microservice fitness platform with Strava/Garmin sync, a Flask ML pace predictor, and Gemini-powered run analysis.
Overview
A running/fitness platform that generates phased workout plans, tracks activities, and gives AI feedback on runs. I built it from the ground up as an intern.
It syncs real data from Strava and Garmin (full OAuth 2.0), predicts training paces with a scikit-learn model, and analyzes runs lap-by-lap with a Google Gemini agent that sends summaries over WhatsApp.
Key Features
Architecture
- Astro/React frontend with 5 NestJS microservices (API + Strava/Garmin connectors) and a Python/Flask ML service.
- 3 PostgreSQL databases (~20 Prisma models), Redis + BullMQ queues, deployed to AWS EC2 via GitHub Actions.
- Exposed as an MCP server so an AI agent can drive the app.
Tech Stack
- Astro
- React
- NestJS
- Node.js
- Python
- Flask
- PostgreSQL
- Prisma
- Redis
- AWS EC2
UPSCash
Cash-reconciliation automation for a US department-store chain — matches each store's end-of-day register cash against QuickBooks and flags the discrepancies that matter. Built in about two weeks.
← ExploreAutonomous AI Testing Tool
DRCODE's main product — agents that test APIs and UIs on their own. I built the UI-testing toolkit from scratch and fixed/improved the API E2E side.
Explore →