Hyfe Cough Tracker - Interview with the developer of a fantastic app

Did you have any prior development or coding experience?

As a serial entrepreneur, I had entrepreneurship and operational experience, not coding experience. However, the other co-founders had coding & software experience so that made for a great combination of complementary skills.

What was the most challenging aspect of developing mobile app?

The app uses AI that has to analyze sound. This means it needs to use the microphone and it has to run in the background. The most difficult was finding a technical solution to this problem that would also be acceptable for our users. As always, the best solution is a combination of well-designed technical architecture, refined user experience, and value created.

_{Iulian Circo}

Iulian Circo specializes in building impact businesses. He has founded and co-founded several global companies and has built a successful track record of entrepreneurship with a focus on exponential technology and impact at scale. He has more than 20 years of experience in building and managing high-performing teams in challenging and extreme operational environments across 4 continents.

Name a few of your favorite apps and reason you love them.

I love what Strava has achieved with the simple use of GPS (and a few other sensors). It is amazing how in a few short years, Strava (and others) have normalized a very precise, data-driven approach to fitness – something that until recently was only available to top professional athletes. Now anyone with a smartphone can train as smart as the best-funded professional athletes.

In fact, the whole health/ lifestyle space gets me really excited. When you think about it, we traditionally know and understand so little about our bodies and our health. Even clinicians and health professionals have resigned themselves to piece together complicated puzzles made out of very subjective interactions with their patients/ clients. How do you feel? Have you been sneezing/ coughing more? How is your headache? These are very generic questions that can hardly be objective data points. Some health professionals are very good at it, while others aren’t. This means that for most people on this planet, there is an enormous barrier of access to basic health services.

Yet, most everyone has a super-computer in their pocket. I get excited by the idea that we could democratize primary health care by simply using the powerful sensors and processing capabilities of smartphones to help generate and interpret real-time, objective data about our health.

This is even more important in these pandemic-ridden days, with social distancing and lockdowns.

How long have you been working on this app?

We started working at the app itself in March 2020. However, the AI work behind it and the clinical work has been years into making.

What need of the user did you have in mind when developing this app?

Ever since we rolled out the app, we discover so many novel ways in which our users use Hyfe:
1. The typical user is someone either with acute or chronic cough. We have a lot of chronic coughers among our users who get value out of finally understanding objective patterns and trends over time for their coughing. We also have many parents who use the app to monitor their child’s coughing – particularly in cases of Asthma, croup cough, and other respiratory conditions.
2. We also have people using the app as a way to track the quality of their environment – for example we saw an uptick in users on the West Coast of the USA during the recent fires – people were tracking their coughs as a way to gauge the seriousness of air pollution in their area. Other people are using it to track their allergies and so on. 3. We have also rolled out a feature that allows people to track the cough of their loved ones – I can easily track my mum’s coughing in real-time, one continent away. We found that quite a few health providers were using this sharing feature in the app as a way to manage patients in social-distancing-compliant ways.
4. Probably the most interesting use of our app is at a population level – if enough people in a certain area are using the app, that allows for essentially a “smoke detector” for Covid19 outbreaks. Even a slight increase in cough frequency at a population level could help municipal public health authority deploy a targeted action and maximize the impact of their limited resources. Here is a short description of how that works

In what way do you think your app is better than similar apps on the market? Please describe in detail what innovation you think you bring and what you are proud of in your app.

There are a lot of worthy efforts out there – usually lead by academics – aimed at collecting as many cough samples as possible in order to train AI models for classification and diagnostic. All of these efforts have a similar model: they encourage people to go to a website/ download an app, tap a button, and then cough three times. This model has a significant weakness: they only collect “elicited” or provoked coughs, which means that we only know if there are any subtle differences between organic/ natural coughs and we have zero context. Do coughs occur in clusters, or randomly? Does a COVID patient cough more in the early hours of the morning or in the evening? All of this important information is lost when we only collect elicited coughs.

Imagine I would want to do a research about happiness and I would go around with a camera asking people to smile in 3, 2, 1…

One of our big innovations is the fact that we focus on collecting naturally occurring coughs. We also collect ALL naturally occurring coughs, along with their timeline context. This puts us in a completely different ballpark in terms of signal to noise ratio as well as in terms of value created for our users (who can identify patterns in their coughing).

What are your future plans and expected features of the coming new versions of this app?

Cough is just one variable for us. Cough is just a data-point – like measuring temperature. What we are trying to build is a continuous, real-time diagnostics streams that not only process random data points but also adjust themselves continuously based on dynamic, continuous data. We are already training models for sneezing, snoring, laughing, crying, and we believe that the most powerful model would consider all these inputs together along with their context, to give our users – as well as their health providers – useful, actionable diagnostics.

Assuming new users of your app are reading this page. What do you want to ask them to do (contact you about X, Share the app, etc.)?

We love to hear from our users. We have learned so much about clever ways in which they have been using our apps and we love to convert the insights from these stories into features.