Building Katha: A Bilingual Phonics Companion with Gemini 2.5 Flash

#ai #gemini #learning #showdev

📚 Katha – Bilingual Phonics & Pronunciation Companion

Built for the Google Gen AI Academy APAC Meet the Builders Campaign.

🌏 The Local Problem

Young learners in bilingual households across the APAC region often lack access to individual speech coaching or localized reading content. Generic tools lack local cultural relevance, making it harder for children to engage.

✨ The Solution

Katha is an interactive web-based educational app that:

Generates 3-sentence, culturally resonant short stories based on a chosen local theme (e.g., Night Markets, regional festivals) in both English and a selected local language.
Acts as an encouraging speech therapist ('Teacher Katha') evaluating real-time student pronunciation inputs against target sentences.

🛠️ Google AI Architecture & Tech Stack

Google Gen AI SDK (google-genai): Interacting seamlessly with production-ready endpoints.
Gemini 2.5 Flash (gemini-2.5-flash):
- Drives the creative story generation module with custom local positioning.
- Powering the evaluation engine utilizing low-temperature system_instruction configurations to provide warm, child-friendly feedback.
Frontend Framework: Streamlit web UI featuring native browser recording overlays.