Inspiration
In the face of a global pandemic, decision makers everywhere need to make tough choices that will affect the lives of millions of people and their loved ones.
However, even though the pandemic is global, decision support is very unevenly distributed.
Some countries lack access to state-of-the-art modelling tools, and many modelling results are not easily accessibly by the general public.
This leaves many decision-makers in the dark, despite the urgent need to make high-stakes decisions.
We are trying to fill this gap.
And in doing so, we want to build a product where model results aren't seen as certain and inevitable; but where we instead convey the uncertainty in the outcome -- and the ability of our choices to change it.
What it does
EpidemicForecasting.org is the most advanced public forecast of COVID-19 developments available for much of the world.
The pipeline consists of:
- Human forecasters estimating true infection rates (thereby correcting for differences in testing capacity etc.)
- Simulations on high-performance computers using GLEAMviz, a state-of-the-art academic pandemic modelling software
- Communicating the results in an intuitive UI that emphasises uncertainty and human agency
How we built it
The tech stack consists in a python backend, bootstrap frontend, and plotly for charting.
Challenges we ran into
- Communicating complex simulations in an accessible manner
- Visualising uncertainty and agency
- Building a pipeline for rapidly delivering modelling results in the face of server constrains on the high-performance computers used for simulation
Accomplishments that we're proud of
- Covering 150+ countries, including many in the developing world, and focusing on decision-makers who wouldn't have accessed this information otherwise
- Shipping an MVP in only two weeks with a newly formed 10+ person team from around the world, including both Oxford academics and volunteer data scientists
What we learned
- How large the gap in decision support is between different decision-makers, and the potential to have a large positive impact by closing it
- The importance of keeping the product lean when moving fast
- The importance of taking ownership and being bold
What's next for EpidemicForecasting.org
How do we defeat COVID-19?
We want to track what methods to reduce transmission work, and how well they work.
To do this, we are building an open-source database of countermeasures, and plan to use machine learning tools to understand it.
The results will then be integrated into our main simulations.
Log in or sign up for Devpost to join the conversation.