Building an Interactive, Offline AI Learning Buddy for My 3-Year-Old Son

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

I built Toddler Learning Buddy, a playful quiz app for my three-year-old son. It turns learning into a simple, interactive game: he can choose a topic, listen to a question and its answer choices, tap an answer, and get encouraging feedback.

The app includes animal and color/shape quizzes, plus counting questions with pictures of objects to count. Counting questions vary the objects and quantities, so each round can feel a little different. I wanted to make it easier and more fun for my son to practice early learning skills with a grown-up nearby.

Demo

The app currently runs locally. To try it, install the dependencies, set up Ollama and Gemma 2B, then run streamlit run app.py.

Code and Video Recording

Github: https://github.com/srinivasskc/toddler-quiz-app

How I Built It

I built the app in Python with Streamlit. It uses Gemma 2B through Ollama to generate child-friendly quiz questions, with inference running locally rather than through a hosted LLM API.

For counting questions, the app randomly chooses an object and quantity, then displays that many matching emoji illustrations and creates answer choices that include the correct count. Questions can also be read aloud using gTTS.

Why Does Open Innovation Matter?

Open models and tools made it possible for me to build a personal learning app without relying on a paid, closed LLM API. Running Gemma locally through Ollama gives me more control over how questions are generated and makes it easier to experiment and tailor the experience to my son.

It also lets me learn from and build on tools that are open and accessible to other developers. The app is a small project, but it shows how open technology can help people create useful, personal experiences for the people they care about.