While searching for a project to tackle, Madeleine shared that her parents are nurses that have to go to work every day, treating patients infected with COVID. She mentioned in particular that both her parents have often expressed frustration with the overwhelming scarcity of personal protective equipment (PPE) for healthcare workers to wear when treating patients during this crisis.

Upon doing more research, we found that this lack of essential supplies extended to food banks as well; many communities around the US do not have consistent access to food from food banks because of COVID. On the flip side, many large corporations like Nike and Ford have been donating and distributing these types of essential supplies (food, PPE). These corporations have the resources to save lives, but we wondered if they knew where exactly to deliver the resources.

We decided to create a platform to inform corporations which areas are in the greatest need of supplies. If we could tell Nike that a particular county in Portland, Oregon was going to be needing significant amounts of PPE in the near future, we could save lives by getting that PPE there as quick as possible. That’s our goal in simplest terms – to help those who need help the most.

What It Does

Our solution is to create an AI tool to help companies prioritize where exactly their donations should be going to. This tool consists of three key features:

1)  Predictive analysis using a linear regression model to forecast the of confirmed COVID-19 cases in a particular county for the next three days, using daily-updated data. Companies can potentially use this predictive model to take preemptive action with their donations and save hospitals valuable time.

2) Graphical visualization of COVID-19 case growth trends over the past week.

3) Visualization of the risk level for all hospitals in a particular county. By dividing the number of confirmed COVID cases in a hospital's city by the number of ICU beds available in that hospital, we calculate a Risk Index that corporations can use to prioritize at-risk hospitals who need supplies the most. The more bottom-right the hospital is on the graph and the lighter its color, the greater its risk.

How We Built It

Tech Stack:

Programming Language: Python

Integrated Development Environment: Spyder

Data Science Tools:  Pandas, SciPy,  NumPy, MatPlotLib

Data:  Microsoft Excel,  GitHub

Front-End:  Marvel UI

Business Tools: Google Docs,  Microsoft PowerPoint, Google Forms

Data Sources: The New York Times, New York State Department of Health

Challenges We Ran Into

None of us had much experience with AI or ML; this was the first major problem that we encountered. We did not have a clear direction or an idea of what exactly we would need to do to get our project completed. We also pivoted on our idea many times as we came across new information. Our mentors were able to help us with everything that we needed, from the tech challenges to the ideation process.

Creating a simple and clear pitch was also a challenge. There is so much information that we wanted to cover, and deciding exactly what to share to create the biggest impact was a challenge that we hadn't encountered before. Once again, our mentors were able to help us quite a bit with this.

Accomplishments We're Proud Of

We are quite proud of getting to where we are at with our project with minimal AI or ML experience coming in -- our linear regression model ended up having an error margin of only .13%!  Even while balancing online school work or exams and working from five different cities, three different time zones, and three different countries, we were still able to find a work schedule that allowed us to be productive and communicate with each other on a regular basis. We all have different backgrounds and different skills, but we were still able to work effectively and produce a valuable product that we are all quite proud of.

What We Learned

This was the first Hackathon for all of us - we each learned how to manage a tech-based project, and dove deeper into how to apply AI and ML to a practical scenario. We learned how to use prototyping tools like Marvel, and how to stay organized when working from many different places in different time zones.

What's Next for A-I Can Help

1) Branching out from PPE for hospitals to other essential supplies, like food for food banks or PPE for nursing homes

2) Marketing ourselves by reaching out to local businesses and large corporations, as well as building a social media presence to attract clients

3) Expanding our database to include information from all counties and hospitals across the U.S., with automatic daily updates

4) Creating an actual website and integrating code for interactive state maps visualizing COVID risk areas.

5) Analyzing the second wave of COVID likely to come in the fall

6) Refining our Risk Index calculation and current predictive model by collecting more data

7) Adapting our tool for use during other, future emergency situations, not just COVID-19

Try It out



github, marvel, matplotlib, numpy, pandas, python, scipy, spyder

Devpost Software Identifier