- My mom is a public health nurse
- Only 50% of her time is spent giving Covid-19 vaccines to people
- The rest of her time is spent calling and collecting information from Covid-19 patients
- What if she spent 90% of her time giving vaccines instead?
- What if she could focus on just the patients that really need her help?
- What if all of her co-workers spent 90% of their time giving vaccines?
- 100,000+ more people would be vaccinated right now
- The Covid-19 crisis would be resolved much sooner
- Less people would be hurt by Covid-19
- How can we use machine learning, automation and the instant communication the mobile phones in everyone's pockets provide to help front-line medical staff keep their patients safe?
- This is the inspiration behind VirusValet
What it does
- VirusValet automates Covid-19 check-ins so public health nurses can focus on giving vaccines and helping the patients that need them the most
- Nurses add Covid-19 patient profiles to VirusValet
- VirusValet automates collecting information from patients by texting them using the Twilio messaging API and asking them questions about their health
- What symptoms do they have?
- Are they self-isolating? etc.
- Using the information it collects, VirusValet highlights patients that need the most help using machine learning
- Nurses can also review the questions the bot has asked the patients and their responses
- Nurses can also ask further questions to clarify the patients' status and give suggestions
How we built it
- We used the Twilio messaging API to send text messages from a Python web application
- We used Google Cloud to speed up the training of our machine learning model and data processing scripts
- Our web application interfaces with the Twilio messaging API and uses the machine learning model to automate communication with patients and nurses
Challenges we ran into
- We had great difficulty finding enough data for our machine learning model
- There were tons of problems that our code ran into when interfacing the machine learning scripts, Twilio messaging API scripts and web application
Accomplishments that we're proud of
- Learning how to use the Twilio messaging API and interfacing it with our machine learning model and web application
- Interfacing three very complex applications together and building a product in a weekend
What we learned
- It is very important to write clean code so it can be easily extended later
- Code should be broken up into different packages, each of which are built properly
- This makes it easier to use them and debug them when problems occur
- A machine learning model is only as good as the data you give it
What's next for VirusValet
- Gathering more data and making our machine learning model more accurate
- Improving our web application to allow nurses to use it easier