Sagar Roy

Software engineer

Sagar Roy

Software engineer

AI-Powered Android Application | Food Recognition | Health & Fitness

See Demo

Pushti AI is an intelligent nutrition and fitness tracking application designed to make healthy living easier, particularly for people who regularly consume traditional Bangladeshi foods.

I independently designed, developed, and published this native Android application, combining artificial intelligence, nutritional data processing, wearable health integration, and personalized weight-management tools into a single platform.

Key Features

AI-Powered Food Recognition

Users can capture food photos and identify meals using an on-device AI model. The application supports Bangladeshi food recognition, Bengali food-name mapping, and server-based AI estimation when local nutrition information is unavailable.

Calorie & Nutrition Tracking

Track daily food consumption, estimated calories, protein, carbohydrates, fat, and other nutritional values. Users can adjust food portions for more relevant nutritional calculations.

Personalized Weight-Loss Goals

Users can set current and target weights, select their preferred weight-loss pace, and choose a target duration. The application dynamically calculates estimated calorie goals and displays daily activity progress.

Health Connect Integration

Integrates with Android Health Connect to retrieve supported activity data, including steps and calories burned, allowing users to monitor their fitness progress.

Progress Monitoring & Reports

Provides an interactive dashboard with circular progress indicators, daily calorie summaries, weight-history tracking, and nutritional reports.

English & Bengali Language Support

Includes multilingual support to make nutrition tracking more accessible to Bangladeshi users.

Technology Stack

  • Android: Native Java, Android SDK, Material Design
  • Machine Learning: TensorFlow Lite / Google AI Edge LiteRT
  • Backend: Python, FastAPI, REST APIs
  • Database: MySQL, SQLite
  • Cloud Deployment: Vercel
  • Integrations: Android Health Connect, AI-based nutrition estimation
  • Architecture: Offline-first food recognition with optional cloud fallback

Technical Challenges & Solutions

One of the key challenges was recognizing traditional Bangladeshi dishes that are less consistently represented in common international food-recognition datasets.

To address this, I implemented a food classification system with localized label mapping, a nutrition database, and an AI-powered fallback for foods unavailable in the local catalog.

Another challenge involved integrating health activity data and dynamically calculating weight-management goals while maintaining a responsive Android interface.

I also addressed Google’s 16 KB memory page-size compatibility requirement by migrating the machine-learning runtime to a compatible LiteRT implementation, enabling successful publication on Google Play.

Project Outcome

Successfully developed and published Pushti AI on the Google Play Store, delivering a functional Android application that combines AI food recognition, nutrition tracking, personalized activity goals, and health-data integration.

The project demonstrates my experience in native Android development, AI/ML integration, backend API design, database architecture, and production application deployment.

Status: Published on Google Play

Google Play: https://play.google.com/store/apps/details?id=com.sagarroy.pushti