GymMate

GymMate centralizes a gym's data and training knowledge in one place. Members follow personalised programs on a mobile app and stay motivated through streaks, XP levels, badges, and a leaderboard. Trainers design programs from templates, assign them, and follow each trainee's progress on a dashboard. Gym owners manage machines, exercises, programs, and users in a web CMS, including a drag-and-drop floorplan that shows where each machine is.

Members get bored, stop seeing results, and lose motivation; trainers juggle programs and trainee history across scattered tools; gym owners need members to keep coming back. A busy or missing machine also breaks a planned workout.

Gym members (mobile), trainers and gym admins (web). The floorplan and equipment data were modelled on CMU Fitness by PlayGround.

Web application developer in a three-person team (web, backend, mobile), CMU Computer Engineering capstone (261492).

I built the web application: the member-facing web pages and the admin/trainer CMS (program builder, floorplan, machines, exercises, muscle atlas, users). Teammates owned the Go backend and the React Native mobile app.

• React
• React Native
• Go (Gin)
• PostgreSQL
• MinIO
• JWT
• OpenAPI / Swagger
• Docker
• GitHub Actions
• Proxmox VE
• Cloudflare Tunnel
• Tailscale
• Dozzle

• Program builder with templates and cloning, so trainers adapt an existing program instead of starting from scratch.
• Exercise substitution: exercises are encoded as feature vectors and ranked by cosine similarity, so the app suggests an alternative when a machine is busy or missing.
• Gamification: streaks, XP levels, badges, challenges, and weekly leaderboards.
• Gym CMS: floorplan editor with drag-and-drop machine placement, plus machines, exercises, muscle atlas, programs, and users.
• Trainer dashboard to assign programs and follow each trainee's progress and history.
• REST API with JWT authentication (HS256, one-hour expiry) and roles for members, trainers, and admins, documented in OpenAPI 3.
• Training logic grounded in the ACSM FITT-VP principle (frequency, intensity, time, type, volume, progression) and self-efficacy research on motivation.

• repositoryProject repository

• Internal and scenario testing, with measured response times, against an experiment design written before testing.
• User survey on the working demo (web app and mobile app).

• The mobile app was not published to the app stores within the project timeline, so no real members used it; evaluation relied on internal and scenario testing.
• The project deployment is no longer online.
• Hardware limits: workouts are logged manually rather than from wearables or machine sensors.

• Writing the API contract (OpenAPI) early let the web and mobile apps be built in parallel with the backend.
• A visible deploy log in chat makes a shared server easier to trust for a whole team.

• 

  Team Projects

  Restaurant POS

  Web point-of-sale for restaurants: customers order by scanning a QR code at the table, the kitchen sees orders live, and cashiers settle bills by table with cash or PromptPay.

  • Next.js
  • TypeScript
  • Tailwind CSS
  • Axios
  • Go (Gin)

  View case study

• 

  Infrastructure

  HomeLab and Local AI

  An old PC rebuilt as a Proxmox server that runs my live projects and a local LLM on its own GPU, published through Cloudflare Tunnel with Grafana and Dozzle for metrics and logs.

  • Proxmox VE
  • LXC
  • Linux
  • Docker Compose
  • Cloudflare Tunnel

  View case study