We were inspired by a problem that goes beyond simply detecting disease: in many rural communities, getting from “something might be wrong” to actually receiving reliable care can be confusing, expensive, and difficult. A patient may not know which doctor they need, whether specialized testing is necessary, or how to navigate appointments, especially when language, distance, and limited medical access are involved. We wanted to make that entire process feel as natural as having a conversation. Instead of expecting patients to understand a complicated healthcare system, our goal became building a personalized voice AI that can communicate in their language, understand their situation, explain the next appropriate step, and help navigate or book care.
Building the project taught us that making the experience simple required a surprisingly complex system underneath. We developed disease specific models using labeled neurological datasets and experimented with preprocessing, feature normalization, class balancing, dimensionality reduction, and multiple machine learning classifiers. For analysis, for example, recordings can be transformed into measurable data such as jitter, shimmer, HNR, RPDE, DFA, and PPE rather than relying on what a person simply sounds like. We connected this analysis with cognitive and ocular information, longitudinal patient history, and a personalized AI layer so the system could use meaningful context while guiding a patient instead of returning an isolated score.
Our biggest challenge was designing advanced technology for people who should never have to understand the technology behind it. Rural deployment can mean different languages, inconsistent connectivity, limited specialist access, and patients who may be unfamiliar with complex medical portals, so we focused on making voice the primary interaction while keeping the intelligence behind the conversation. The AI is designed to turn complicated results and patient context into understandable guidance and help move someone toward appropriate care, including navigating and booking appointments when needed. Ultimately, we learned that the most valuable part of the project was not making another health result database, but building a bridge between sophisticated neurological technology and people who might otherwise struggle to access it and helping them solve a problem, not just discover it.
Log in or sign up for Devpost to join the conversation.