heal.io
Dec 2019 - Jan 2020
About
heal.io was an AI-powered health assistant built to help with everyday health issues. It worked a bit like Siri, but focused entirely on healthcare. You just had to type in your symptoms, answer a few quick questions, and within seconds, it would suggest a likely diagnosis along with simple home remedies.
It could also find nearby doctors and help you book an appointment. heal.io supported 24 common health conditions, making it a fast, accessible tool for users without stable internet access.
Why build heal.io?
During our research, we found that the average Indian visits a doctor 8 to 9 times a year, despite a poor doctor-to-patient ratio of 1:1000. Most patients end up waiting close to twenty minutes for appointments that last less than two.
It made us wonder if that kind of waiting is really necessary for minor issues. That is where the idea for heal.io came in. We wanted to give people quick, reliable health advice right when they needed it.
How it works?
heal.io worked as a simple tool where users could list their symptoms directly, without needing to chat with the system. Based on the input, it used a trained model to predict the most likely condition and suggest home remedies. Users could choose between a voice output mode or a text-based response, depending on their preference.
Technical specs
This project was built using Python and JSON, supported by several libraries. We used NumPy for mathematical operations and array handling, and the Random module to generate dynamic responses. For natural language processing, we implemented NLTK (Natural Language Toolkit), which enabled the chatbot to understand and respond in a more human-like manner. To provide audio feedback, we used gTTS (Google Text-to-Speech), allowing the assistant to deliver spoken responses. At the core of the system was TensorFlow, which powered the machine learning logic behind the symptom-to-condition mapping.
All health conditions and their mapped responses were stored in structured JSON files. These were processed by the TensorFlow model in real time.