CureAI

Inspiration

Healthcare is a basic human right, yet millions lack timely access due to cost, distance, or shortage of doctors. We wanted to create something that could bridge this gap and empower people with reliable, immediate health guidance, right at their fingertips.

What it does

CureAI is an AI-powered health assistant that allows users to describe their symptoms in natural language. The system provides:

Possible explanations for their condition
Safe home remedies when appropriate
Clear advice on when to consult a doctor or seek emergency care
It’s not a replacement for medical professionals, but a supportive tool that reduces uncertainty and helps people make informed decisions.

How we built it

Used OpenAI’s gpt-oss models (120B) as the reasoning core
React for the frontend
FastAPI for the backend
Integrated via Hugging Face for inference
Built a simple web/app interface where users can enter symptoms
Designed prompt structures to ensure safety-first, context-aware outputs
Optimized responses to prioritize accessibility, empathy, and clarity

Challenges we ran into

Balancing accuracy with safety — ensuring the AI doesn’t overstep into providing unsafe medical advice
Optimizing latency and inference speed for large models locally
Designing responses that feel human, empathetic, and clear rather than purely technical
Keeping the system useful offline or in low-connectivity environments

Accomplishments that we're proud of

Built a working prototype that provides safe, context-aware health advice
Successfully applied open-weight reasoning models for a real-world humanity-centered use case
Designed an intuitive interface that makes the technology accessible to anyone, not just tech-savvy users

What we learned

How to effectively prompt and guide large reasoning models in sensitive domains like healthcare
The importance of human-centered design when working with AI that touches people’s lives
Strategies for running large models locally and optimizing inference for practical use cases

What's next for CureAI

Expand CureAI into a multilingual assistant for global accessibility
Add speech-to-text and text-to-speech to support users with low literacy
Work with healthcare professionals to refine safety guardrails and accuracy
Explore deployment on edge devices for offline usage in rural or remote areas

Built With

fastapi
gpt-oss-120b
huggingface
openai
react

Submitted to

OpenAI Open Model Hackathon

Created by

I was responsible for building CureAI from the ground up during the OpenAI Open Model Hackathon. My key contributions include:

Idea & Design – conceptualized CureAI as a healthcare assistant under the For Humanity category to improve accessibility to health knowledge.

Model Integration – integrated OpenAI’s gpt-oss models (20B & 120B) using Hugging Face for inference, and engineered prompts to ensure clarity, empathy, and safety in health responses.

Backend Development – built the backend using Python + FastAPI, enabling smooth communication between the model and user interface.

Frontend Development – implemented a simple React + Tailwind interface for users to easily enter symptoms and view guidance.

Testing & Validation – created sample input/output cases to validate that CureAI prioritizes safety and provides reliable suggestions.

Documentation & Submission – wrote the README, testing instructions, and structured the Devpost submission to clearly communicate the project.

In short, I contributed to all stages of the project — from ideation and coding to testing and documentation — ensuring CureAI is not just functional, but impactful for humanity.

Muhammad Taha