Project Info
Inspiration
Healthcare workers are generally very stressed and the technology used in many hospitals and healthcare firms are very outdated. One of my friends told me about how in the hospital he works at, they still page doctors instead of just instantly messaging them, causing inefficient doctor transfers and slower communication. And many of these inefficiencies transfer over to the patients, who are usually anxious waiting for answers from the doctor. Fast Aid PI tackles a specific area: patient prediagnoses. We hope to relieve stress for the patient, give them a course of action before their doctor appointments, and speed up data transfer from patient to doctor.
What it does
Our app takes in medical data from the patient, prediagnoses them using a machine learning model (mocked by an LLM API call), gives them a recommended path of action, and transfers the data to the doctor to take notes for their upcoming appointments.
How we built it
For the planning phase, we consulted friends who worked in the medical space and asked for any large problems or concerns they experience. We built it using a node.js frontend and fastAPI backend server.
Challenges we ran into
We had various challenges while prototyping the idea for this app: How to stay within HIPAA guidelines for sharing information and data How to prevent users from getting overwhelmed by information online How to ensure users effectively document their well-being to the doctor How to make sure doctors can see and view information effectively As for technical challenges, the main challenge came via communicating clear goals and linking the client and server architectures.
Accomplishments we're proud of
Following the structure of the challenges above, here is how we tackled each one of them: Via providing an easily scalable authentication framework with password encryption, preventing views from users and practitioners unless users give consent Providing a simple, easily followable structure given in the prediagnosis phase rather than listing out all possible medications. We want the actual diagnosis responsibility to be the doctor's. Providing an AI-powered chatbot to help ease potential stress induced on the user and clarify any questions before going to a practitioner Providing persistence of chats, forms, and prediagnoses on the user side Providing a summary view of user prediagnoses that doctors can take notes on
What we learned
We learned a lot about the problems in healthcare, prototyped various ideas, and decided to tackle prediagnosis and appointment stress. Through researching problems and communicating with people working and/or studying healthcare, we were able to determine the best course of action for each feature of our app.
What's next
There are a lot of improvements that can be made! We have some stretch features that we were not able to get to and some bigger issues that we couldn't tackle: Doctors should be able to make edits and highlight certain parts of prediagnosis results sent from patients. Adding a system where doctors can easily share patient diagnoses and treatment with one another. This tackles the communication error so common when patients must deal with multiple specialists to treat their cause. Users registered as doctors can add a patient medical history to preprocess the AI-powered prediagnosis step to create a more accurate report. Enforcing stronger authentication and encryption guidelines to abide by HIPAA guidelines. Training an actual ML model based on real patient cases and doctor prescriptions to improve runtime and cost efficiencies. Constantly calling a powerful LLM is unnecessary for the prediagnosis of symptoms and potential diseases.
Fast Aid PI
Bridging the gap between patients and doctors through AI-powered prediagnosis and communication. Various features to help prediagnose patients, offer solutions, refer to specialists and relevant practitioners, alongside features to help doctors provide care more efficiently.
Features:
- Secure authentication and role-based access
- AI chatbot for prediagnosis and patient reassurance
- Prediagnosis Feature: Analyzes questionnaire then offers action plan, recommended practitioners, and health support.
- Persistent chat and prediagnosis records
- Patients can add relevant recommended doctors to their conversations and prediagnoses.
Tech Stack:
- Frontend: Typescript, React, Tailwind, ESLint, Anthropic
- Backend: Python, FastAPI, uvicorn, SQLAlchemy, Anthropic, pyjwt, passlib,
Getting Started:
Backend: Uses uv to handle packages.
$ cd fast-aid-pi/server
$ uv install
$ uv run server.py
Server should be running on http://localhost:8000
Frontend: Packages managed by npm.
$ cd fast-aid-pi/server
$ npm install
$ npm run dev
Frontend components should be running and visible on http://localhost:3000
Next Steps:
- Allow doctors to edit and highlight patient summaries
- Enable cross-specialist communication for holistic treatment
- Add patient history integration for more accurate prediagnoses
- Improve encryption and compliance
- Train a custom ML model for prediagnosis instead of relying on an LLM
Analysis
View
Metric
- 13
- 10
Figures cover GitHub contributors during the hackathon window. A co-authored commit counts in full for each author, so per-member totals add up to more than the whole-team figures.
Technology
- AnthropicIn code
- CSSIn code
- FastAPIIn code
- JavaScriptIn code
- Next.jsIn code
- PythonIn code
- ReactIn code
- Tailwind CSSIn code
- TypeScriptIn code
- Node.jsClaimed
9 of 10 appear in the indexed code. 1 claimed on Devpost could not be matched to code, which may simply mean the tool leaves no trace in the repository.
AI coding agents
No AI coding agent signals were found in this repository.
Detected from committed agent config files and commit authorship. Absence of a signal is not proof an agent was unused.
Codebase size
Source size
161 KB
Source files
30
Counts recognized source files only; vendored directories, binaries and lockfiles are excluded, so this is smaller than the repository on disk.
Repository
ericqn/fast-aid-pi
46 files · 601 KB · @ 75d754d
Structure
Interface
10 files · 22%Screens, components and styles rendered to the user.
API & routing
15 files · 33%Request entry points: routes, handlers and controllers.
Data & schema
4 files · 9%Schema definitions, migrations and data access.
Supporting
Layers are inferred from where files sit in the tree, not from reading the code. A project that names its directories unconventionally will read oddly here — open the file browser to check anything the diagram implies.
Languages
- TypeScript67%
- Python29%
- Markdown2%
- Shell1%
- CSS0%
- JavaScript0%
Share of indexed source by file size. Binary and vendored files are excluded.
Dependencies
client/package.json
npm · 12- @anthropic-ai/sdk
- next
- react
- react-dom
- +8 more
server/pyproject.toml
pypi · 11- anthropic
- bcrypt
- email-validator
- fastapi
- passlib
- pyjwt
- python-dotenv
- python-multipart
- requests
- sqlalchemy
- uvicorn
Declared in the repository’s manifests at the indexed commit. A declared package is not proof it is used, and runtime dependencies are listed first.
This project’s features have not been analysed yet.
Export this project's context (description, README, evidence, key source files) to chat with an AI agent elsewhere.