SweetAudio

We are trying to solve the challenges of type 2 diabetes in underrepresented population & low income countries. There are over 530  millions of people suffering from Diabetes Type 2 which is 10.5% of global population. Also, 4 out of 5 (81%) diabetes patients are from low- and middle- income countries. Over 70% of patients people lack access to regular monitoring of glucose due to the cost and accessibility of devices (traditional glucometer & CGMs). 

An estimated 44% of adults living with diabetes (240 million people) are undiagnosed. Almost 90% of these people live in low income and middle-income countries. Diabetes was responsible for an estimated USD 966 billion in global health expenditure in 2021.

Since lack of regular monitoring can lead to chronic conditions such as diabetic foot, ulcers, chronic kidney disease and heart disease etc, there is a huge problem in underserved populations and low/middle income countries especially when health expenditures are out of pocket.

SweetAudio is a technology that can measure your blood glucose level using your voice. So, whenever the glucose level changes, the pathopysiology changes as well. First, the elasticity of vocal cords & larynx changes. Second, in hyperglycemia - there is anxiety and urge to speak faster & in hypoglycemia - there is lethargy and slurred speech as well. Based on this medical background, we build our machine learning model using both voice data & glucose data. 

So, you input your voice data and glucose data (from traditional glucometer or CGM) for two weeks. Since everyone has different voice, this personalization is important to build the baseline model. Current models can predict up to 89% accuracy. Later, you can boost the accuracy and also predict the glucose level for next few hours by inputing lifestyle data & food data as well.

So, we want to give this solution's free version to the underserved population and people from low income population. This free version will be supported from Ads revenue in app and also we can collect more data if we can do faster user adoption. Patients who cannot afford CGM or traditional glucometer can easily measure their glucose level with their phones.

Second, we will target premium version to people with fitness goals. We understand that fitness population has high interest for their glucose and insulin level for optimizing their workout and diet.

Once we got enough data and budget, we want to launch it as medical use. This needs more trials and research. We want to go through FDA at later stage for rolling out medical use. For this, we will consider insurance payment pathway.

Each member of our diverse team brings a unique perspective shaped by their backgrounds and experiences - ML Engineer, Medical Doctor, MBA Candidate/Scientist and Public Health Professional. We all are Asian immigrants and 3/4 of the team are women. One of the members is refugee. From designers to developers, many of us have roots in communities facing similar healthcare disparities, and the team really understands the needs of the targeted low-income population. We have close ties with some of the NGOs in the Asia region so that any further studies can be done in those low income countries.

The shared experiences between our team fuel our passion for creating solutions that truly meet the needs of those we serve. Moreover, our approach to design and implementation is deeply rooted in community input. We believe in co-creation, working hand in hand with community members to understand their needs, aspirations, and challenges. Through extensive outreach efforts, including focus groups, surveys, and community meetings, we ensure that the voices of those directly impacted by our solution are heard at every stage of development. This collaborative process not only ensures that our solution addresses real-world needs but also fosters a sense of ownership and empowerment within the community.

Runa Khan, who is a public health advocate, is deeply immersed in addressing healthcare disparities. She brings first hand experience among the diabetes population in Pakistan to our initiative aimed at revolutionizing diabetes management through voice recognition technology. Her journey, marked by a Master of Public Health degree, a public health consultant at the WHO and ongoing studies in Epidemiology at Harvard Chan School of Public Health, reflects her commitment to advancing equitable healthcare solutions.

Caiwei Tian brings not only her expertise as an ML engineer and healthcare scientist but also a deeply personal understanding of the challenges faced by underserved communities in accessing healthcare. Growing up in a community where one doctor struggled to serve thousands in China, Caiwei witnessed firsthand the repercussions of limited medical resources. This personal connection drives her commitment to designing and leveraging everyday devices to monitor health conditions, providing convenience to patients with chronic diseases, and alleviating their financial burdens.

We have built our own prototype (Machine Learning Model) and UI of the app. About 10 participants' voices were recorded during the Ramandan period before and after fasting. Our target is up to 1,000 participants to improve our baseline model.

We want more connections for our anticipating roll-out plans. We want to build more partnerships with NGOs and research institutes from the networks of Solve. Moreover, we want mentorship for our go-to-market strategy and long term FDA Approval as well.

Traditionally, the glucose level is measured by testing blood, which is inevitably invasive and causes pain. Image patients have to do it multiple times everyday for the entire life. 

Our solution, empowered by AI, finds an association between voice features and blood sugar level. Using the external features, not only the cost will be significantly reduced, it is a lot more convenient to use with no pain at all. We also provide tools to predict blood sugar level over the next few hours to reduce the frequency they have to worry about testing. 

There is no comparable product in the market right now. Diabetes patients have to wear a device all the time or to take a blood test to measure their glucose, which can be a huge burden both physically and financially. Our AI product, the non-invasive, easy, and cheaper solution to measure blood sugar, will significantly improve their life quality and reduce the risk of complications.

We desire to build an AI model to predict glucose level using voice. There are two core steps in our technology. 

First, we need to process the audio data. When user speak to their phone, we will record the audio and process it to grab voice features, such as number of fundamental periods, fundamental frequency, energy, relative average perturbation, shimmer, F1 to F4 harmonic to all energy ratio, etc..

Second, with the voice features we get, 

Four People

1 Full time & 3 Part time

We have been working this solution over 3 weeks since we won the hackathon ar Harvard.

We really understand about DEI values. 3/4 of the members are women and one of the members is a refugee. Everyone is immigrant in the US and 2/4 come from low income country. All of us have more or less faced the problem of inequity and we fully understand how hard it can be without support from the community. We highly value a diverse, equitable, and inclusive environment and we are facilitating that by building a team with various background. Everyone is valued and respected in our team. Everyone's unique experience will also contribute to us providing products that further improve diversity, equity, and inclusion in the community.

There are two tiers of our product. 

For free version, we provide real-time glucose level measurement. The revenue comes from Ads in app. Possible Ads are coupon website, local market, etc., targeting low-income population.

For premium version with subscription, we provide more services. Users could get a prediction of their glucoe level over the next few hours, so that they can be better prepared to take enough insulin if they are going out. Plus, we will ask users to input their meal information. By analyzing their sugar intake, calories intake, and their daily activities, we will provide diet and exercise suggestions for them to maintain a stable glucose level. Subscription model will target group of lifestyle active population, or pre-diabetic population, who care about their blood sugar level and want a healthier lifestyle.

In the future, if we could get FDA approval, we can get insurance payment to use our product as a medical device. Compare to other test methods (CGM, blood test, etc.), out product has a much lower price. 

Apart from our current revenue model, we will also work together with Research Institutes & NGOs and try for grants of the research.