Traveling is hard (or impossible) due to coronavirus. Though it is not suggested that you go out, there may be certain reasons you have to go out. But we want to make sure that the place you are going is safe to visit on a specific day. This app lets you know ahead of time what the case counts might look like for a particular day at a particular place.
What it does
The project started as a model for predicting the number of COVID cases in the coming week. Eventually, we ended up with a web portal that lists the expected COVID stats for each district and a mobile application, which helps its users to plan travel accordingly. It tells whether a person should go out to a particular place at a particular time. The models are updated daily so things might change based on the daily data.
How we built it
We built the backend using Julia. For predictive modeling, we took inspiration from Universal Differential Equations and used a multilayer perceptron to parameterize the differential equations, and using backpropagation through the solver we are able to learn the parameters in an end-to-end manner. For the android app, we used react-native.
Predictive Modeling of COVID-19
We used DifferentialEquations.jl to model the SIR differential equation model. We used a small 2 layered neural network to parameterize the differential equation. The modeling was inspired by  and . We obtained the daily active cases and resolved cases for each Indian district, and accumulated the data since the day the first 10 cases were reported in each district.
Embedding a differential equation allowed us to compensate for the lack of a large quantity of data by building on top of a strong prior. Additionally, since the data is noisy due to improper testing and backlog of test results, we give more weightage to more recent data. Finally, we are able to train the model in an end-to-end manner by minimizing the weighted L1 norm of the difference in the predicted and actual active cases.
 Rackauckas, Christopher, et al. "Universal Differential Equations for Scientific Machine Learning." arXiv preprint arXiv:2001.04385 (2020).
 Dandekar, Raj, and George Barbastathis. "Quantifying the effect of quarantine control in Covid-19 infectious spread using machine learning." medRxiv (2020).