Case study · 2026
English Learning Application for Early Childhood (ELFEC)
An edge-cloud Android application developed for early childhood English education. Features interactive vocabulary learning, handwriting recognition via Google ML Kit OCR, offline pronunciation detection via Sherpa-ONNX, and local multiplayer minigames via Google Nearby Connections.
Project gallery
Splash screen
Overview
Developed an interactive Android-based English learning application for early childhood, combining vocabulary acquisition, handwriting exercises, pronunciation evaluation, gamification (XP, levels, leaderboards), and single-player as well as local peer-to-peer multiplayer minigames.
Problem & goals
Early childhood learners need engaging, interactive feedback to practice language skills effectively. However, many existing educational tools require constant internet access or lack real-time validation for written characters and pronunciation.
My contribution
Architected the application with Kotlin and Jetpack Compose. Integrated on-device Google ML Kit OCR for handwriting recognition and Sherpa-ONNX for offline speech evaluation. Integrated Groq API generative AI for bilingual contextual examples, built Firebase authentication and progress tracking, and developed local multiplayer communication using Google Nearby Connections.
Process & technical decisions
Designed a hybrid edge-cloud architecture where latency-sensitive OCR and speech processing run strictly on-device to ensure zero latency and offline availability, while cloud APIs provide contextual content. Managed the repository with a strict Git workflow using feature/fix branches, pull requests into dev, and Conventional Commits.
Results & lessons learned
Successfully delivered a responsive learning application with real-time on-device evaluation, offline accessibility, and local multiplayer connectivity, fulfilling the requirements for the undergraduate thesis with distinction.