ritwikkanodia / intuition-v7.0

VaccineTogether is an AI-enabled solution for predicting vaccine side-effects before a dose and global data crowdsourcing to perform analytics, all without compromising user privacy & anonymity.

VaccineTogether

Imagine being able to know the side-effects YOU might have before being administered a COVID vaccine. Presenting VaccineTogether, a one-stop solution enabling individuals to be more informed and confident before getting a COVID Vaccine. Moreover, it will enable health care professionals and researchers to crowdsource data, without compromising data privacy and anonymity.

Problem Statement

Our application intelligently addresses two problem statement under the SAP Cloud Track, which are as follows:

1. Smart data sharing: How can various countries share data while preserving the privacy of the public in order to order to understand the virus and its effects better?
2. Vaccination Effects: How can we better track and monitors the effects of vaccinations on administered people in order to understand their effects and improve them further.

What it does

The Application offers 3 key functionalities:

1. Predict side-effects of a vaccine for a specific user based on their health-related data, while preserving user privacy and anonymity
2. Enable users to track their vaccine side-effects using useful analytics as well as crowdsource data for improving the accuracy of the central ML model.
3. Provide global data analytics and insights about vaccination side effects

How we built it

The pipeline is explained by the diagram above. The communication between the local model and the global mode is made with the help of 2 Flask APIs in the backend. Firstly, the Global Model API fetches the model weights, which is hosted on the SAP Cloud Platform. Secondly, another Flask API calls the local model, initializes the weights from the Global Model and finally returns the predictions.

Tech Stack:
Machine Learning: Tensorflow, Keras.
Web App: JavaScript, Flask, Python.
UI: SAP UI5 Framework based on Fiori Guidelines
Maintaining User Privacy and Anonymity: Using Federated Horizontal Learning & Deep Neural Network.

Challenges we ran into

• Finding the right dataset to train our models
• The datasets are very recent (1 Month old) and hence with no documentation or metadata
• Data Preprocessing, cleaning and extracting useful insights
• Integrating the Central Machine Learning model hosted on the cloud, to the locally hosted machine learning model (Updating the model weights with the new gradients. Refer to horizontal federated learning)

Accomplishments that we're proud of

• Predicting side-effects on new users, with high accuracy for the federated model (Compared to a traditional model trained on a completely centralised dataset)
• Training the model comprising a user's personal information, and health data, through horizontal federated learning
• Deploying an entire ML-based application to the SAP Cloud Platform with client-side integrations
• Understanding about the state-of-the-art method Horizontal Federated

What we learned

• We learned how to use the SAP UI5 framework for building responsive application following the Fiori guidelines.
• We deployed our entire application on the SAP Cloud Platform, leveraging its robust infrastructure and platform for application development
• Staying awake for 24 hours (I am writing this at the 23rd hour of the Hackathon XD)

What's next for VaccineTogether

• Increase the acceptance rate of our application to get more data
• Improving our central Federated ML Model as and when more data is available from our users.
• Vaccine Cold chain integration with SAP Ariba and SAP S4HANA.

Screenshots (UI5 App)

Full Presentation

Feel free to watch the full presentation to get a more in-depth overview of our product. The trimmed 4-min version is available at the top of our DevPost page.

Team Photo

Installation

Steps:

1. Download/ Clone this git repository

To run the Juptyer notebooks, download:

1. Download Jupyter notebook
2. Navigate to the notebooks directory
3. Run each cell

To run the UI5 Web App Application:

1. Execute ui5 serve -o /index.html

To run flask backend:

1. Execute FLASK_APP=model.py flask run

The global model is already hosted on the SAP Cloud Platform.

Team Members

1. Jay Gupta
2. Ritwik Kanodia
3. Palak Somani
4. Aditya Bansal

Winners of iNTUition v7.0 in SAP's Cloud Track.