Project spotlight

CharacterGuesser

A browser-based handwriting game where users draw characters and a TensorFlow.js model predicts them in real time.

JavaScript TensorFlow.js Static Web App
CharacterGuesser preview ML Canvas
Launch live demo

Opens the deployed static site in a new tab.

What it is

CharacterGuesser is a lightweight, static web app that turns handwriting recognition into an interactive game. Users draw directly in the browser, then watch predictions update as they refine their strokes.

Behind the scenes

  • Multiple CNN models are trained in Python and exported to TensorFlow.js format.
  • The static app bundles the model with HTML, CSS, and JavaScript for zero-backend hosting.
  • A dedicated game mode challenges users with timed rounds and score tracking.
  • Deployment is streamlined through a docs build step for GitHub Pages.

Why it matters

This project showcases end-to-end ML delivery: model training, export, and a polished client-side experience. It highlights product thinking around responsiveness, clarity, and keeping UX fast without server dependencies.