🗄️ Archived college project — built to learn Android development with Kotlin. Not actively maintained; kept public as a record of early mobile-dev work.
A Kotlin-based Android messaging app with ML-powered Smart Reply — built as a college project to learn native Android development, Firebase, and on-device NLP.
Connect is a real-time chat app for Android with a Smart Reply feature powered by Firebase ML Kit's on-device NLP model — it analyzes recent conversation history and surfaces up to 3 contextual, one-tap reply suggestions, the same pattern used by Gmail/LinkedIn quick-replies. The goal was hands-on learning: Kotlin + the Android SDK, Firebase as a real backend, and integrating a pretrained ML model into a mobile app end-to-end.
Native Android client (Kotlin) with a Firebase backend for auth and real-time message sync between users. Each conversation's recent message history is passed to FirebaseNaturalLanguage.getInstance().getSmartReply() (Firebase ML Kit's on-device Smart Reply model), which returns contextual reply suggestions the user can tap instead of typing.
git clone https://github.com/rk1091/Connect.git
# Open in Android Studio
# Add your own google-services.json (Firebase project config) to /app
# Build and run on an emulator or device- First hands-on experience with Kotlin and the Android app lifecycle, coming from a mostly web/backend background
- Wiring up Firebase for auth and real-time data sync — an early, practical intro to working with a managed backend service rather than owning the full stack
- Integrating a pretrained ML model (Firebase ML Kit Smart Reply) into a live app — passing conversation state into an inference call and handling the async suggestion result in the UI
- Working in a team codebase on a mobile project, distinct from the solo web projects I'd built until then
College team project, built to learn Kotlin/Android. Kept public as part of my project history, not representative of my current (backend/cloud-focused) work.