AURA: Offline AI Diagnostics for Malaria Detection

Light mode
Dark mode

Inspiration

In many parts of Africa, healthcare delivery is hindered by limited access to diagnostic tools, unstable internet, and power constraints. Inspired by these realities, we built AURA to bring offline AI-powered malaria diagnosis to smartphones. Our goal: enable frontline health workers to make rapid, reliable diagnoses—even in the most remote areas.

What It Does

AURA allows users to snap or upload a malaria test slide image using a smartphone, then runs a lightweight AI model locally to detect the presence of malaria. Results are saved offline for later review — with no internet required.

How We Built It

Developed using Kotlin (Android) and Android Jetpack libraries
Integrated CameraX for live image capture
Built a Convolutional Neural Network (CNN) for malaria cell detection
Converted and optimized the model using TensorFlow Lite for offline inference
Used Room Database for storing diagnostic records offline
Designed the app UI with Material Design principles for simplicity and clarity
Optimized and tested performance on low-end Android devices

Challenges We Faced

Optimizing the AI model to ensure smooth offline performance on low-resource devices
Managing camera quality variance across Android hardware
Ensuring data accuracy and feedback despite compute limitations
Creating an intuitive experience for non-specialist health workers

What We Learned

Practical methods for delivering AI solutions in resource-constrained environments
How to optimize deep learning models for mobile inference
The power of offline-first thinking when designing for underserved regions
The value of interdisciplinary design — combining healthcare, AI, and mobile UX