SwasthAI

InspirationMillions of people in rural and underserved communities still lack timely access to quality healthcare. Frontline health workers, such as ASHA workers and paramedics, often have to make critical decisions without adequate diagnostic support. This healthcare gap inspired us to build SwasthAI—an AI-powered point-of-care triage assistant that empowers frontline workers with intelligent decision support, enabling faster diagnosis, better referrals, and improved patient outcomes.

What it does

SwasthAI assists frontline healthcare workers by collecting patient symptoms and vital signs, analyzing them using AI, and suggesting probable conditions with an urgency level (Green, Yellow, or Red). It also supports image-based screening for visible conditions such as skin diseases and wound infections, works offline for low-connectivity regions, and automatically prepares referral information for critical cases. Rather than replacing doctors, SwasthAI acts as a force multiplier, helping healthcare workers make more informed decisions.

How we built it

Our solution is designed as an offline-first mobile application powered by lightweight AI models.

Symptom triage engine using machine learning/LLM-assisted reasoning trained on open medical datasets.
Image screening module using transfer learning with CNN architectures such as EfficientNet or ResNet.
Offline-first architecture with local storage and background synchronization for unreliable internet connectivity.
Smart referral system that routes critical patients to the nearest appropriate healthcare facility with structured patient information.

The solution is designed to be scalable, affordable, and deployable on low-cost Android devices commonly used in rural healthcare settings.

Challenges we faced

Building a healthcare AI solution presented several challenges:

Ensuring clinically meaningful recommendations while avoiding over-reliance on AI.
Designing an interface simple enough for minimally trained health workers.
Planning for offline functionality without compromising usability.
Balancing accuracy, speed, and computational efficiency on low-end smartphones.
Addressing ethical concerns such as bias, explainability, and responsible AI usage in healthcare.

What we learned

Working on SwasthAI taught us that impactful healthcare innovation goes beyond building accurate AI models. Success depends on understanding real-world constraints such as infrastructure, user trust, accessibility, and clinical workflows. We also gained valuable experience in AI model selection, mobile system design, healthcare sustainability, and human-centered product development.

Impact

By enabling faster triage at the point of care, SwasthAI aims to reduce unnecessary hospital referrals, lower patient travel costs, improve early disease detection, and strengthen healthcare access in underserved communities. Our long-term vision is to support government health programs and contribute to equitable, technology-enabled healthcare for everyone.