Replica Health Is Simplifying Type 1 Diabetes Management with Conversational AI
Sam Royston and Brad Jacobson are trailblazing a new era in Type 1 diabetes management with their innovative platform. The Replica Health app will stand apart as an intuitive personal assistant, harnessing the power of conversational AI and natural language processing to offer actionable insight to users’ personalized queries in plain English.
Investors, learn how you can back Health Transformers like the Replica Health team.
Challenge
Imagine walking on a tightrope, carefully balancing between two towering skyscrapers. You take each step with precision and unwavering focus. Now imagine doing the same blindfolded, unable to see where you’re going. All you can rely on are the numerous voices shouting contradictory directions. That’s what it feels like to be a Type 1 diabetic wearing continuous glucose monitors (CGMs) and insulin pumps. You get constantly bombarded with data but are starved for actionable insights.
As you stand atop the dizzying wire, you are handed a torrent of numbers, charts, and alarms from the devices that help you keep your blood glucose levels in check. Glucose readings, trend lines, insulin doses, carbohydrate intakes — it’s like drowning in an ocean of numbers, each representing your health. Suddenly, you realize that despite all the impressive technology at your disposal, you still lack the clarity and insights you desperately need.
Enter Replica Health, a wise personal assistant who analyzes all those data on your behalf and gives you answers in plain English.
Origin Story
Like so many founders in the diabetes market, Sam Royston’s desire to innovate came from a personal place. He was diagnosed with Type 1 diabetes at age 10 and then built a career as an expert in machine learning and data science. Year after year, as he grew in his knowledge, he sought ways to better manage his condition.
During his graduate studies at NYU’s Courant Institute in 2016, Sam began working on Type 1 diabetes directly. His work centered on predicting blood glucose levels and developing new algorithms to enhance these predictions. He also sought to leverage new data sources to improve the accuracy of blood glucose forecasts. This earlier research was the starting point for Replica’s technology.
Then, In 2017 Sam helped start an anomaly detection platform called VoteShield, which uses ML to make sense of civic datasets. That work brought his thinking to a new level, and introduced him to other collaborators with a similar passion for T1D.
What Sam came to realize was that while diabetes management had improved over the years, especially with the advent of CGMs, it was still lacking in several critical areas. The most important of which was how people who wear insulin pumps and CGMs were creating tons of data but had a hard time correlating that with appropriate actions.
Unlike Type 2 diabetics, people with Type 1 diabetes generate a vast amount of data in their daily lives, as they have their blood sugar levels and insulin dosage constantly monitored. However, insulin pump data was often overlooked in retrospective analyses. The combination of insulin dosing and blood glucose data holds immense potential in understanding the carbohydrate release profile of various foods, the effects of various activities, and other metabolic questions.
The existing food-tracking software lacked the essential longitudinal analysis required to comprehend the full impact of a meal over time. A good understanding of the carbohydrate release profile demanded deeper insights into the timing and dynamic component of each individual’s data. And these picture-based apps, even with volumetric scanning, could only provide limited information, often reduced to scalar values like the number of carbs in a particular meal.
Sam and collaborator Brad Jacobson reached the conclusion that the key to better Type 1 diabetes management was to develop a method that incorporated all available time-based data, allowing for a comprehensive understanding of the metabolic responses to various meals, activities, and life events. In addition, they had to create a simple and intuitive solution, allowing people with Type 1 diabetes to gain insights into their conditions and metabolic changes without having to constantly spend their time on data entry or finding their way around technical, confusing platforms.
A potential roadblock to developing the solution was the dynamic nature of data which had to be interpreted; often drawing from multiple sources. Using learnings from VoteShield, where a massive pipeline of evolving time-series data posed similar challenges, Replica overcame these issues quickly.
“When you have heterogeneous data sources like blood glucose and activity data from patients that should be brought into your platform, they should be constantly synchronized. The databases can shift under your feet, and it’s important to keep track of these changes to generate accurate outputs,” says Sam.
With every round of designing, Sam and Brad focused on making their platform’s interface simpler and simpler. They wanted to create a platform that would tell individuals using CGMs what they did right and where they went wrong. This way, they could leverage these insights to make informed decisions the next time they have a similar meal without excessively interacting with the app.
The result of their collaboration was Replica Health, an innovative platform that amalgamates data from multiple sources, including blood glucose readings and location information, to provide insightful reports and predictions. The app would fill a void in the market by offering a unique natural language search interface, allowing users to query their data in plain English and receive detailed reports tailored to their questions.
Today, the duo is fine-tuning the app and working on numerous features to be added to the platform soon. They are also just weeks away from submitting their app to the app store.
Under the Hood
Replica Health is a user-friendly and intuitive app that utilizes data from CGMs and insulin pumps to create a feedback loop for users. It leverages conversational AI/natural language processing to decipher users’ complex and personalized questions about Type 1 diabetes and provide relevant answers based on their historical data. In other words, a user can query the app in English and get answers to their most pressing questions about their condition.
To explain how this works, Brad gives examples of two prototypical questions a user can ask the app.
“A user could ask the app, ‘What happens to my insulin sensitivity when I exercise on vacation?’ In response, the app would analyze relevant data, such as the user’s location, calorie burn (measured on the Apple Watch), the type of exercise, etc., before showing them their insulin sensitivity for those occurrences. Similarly, another question could be, ‘Show me all the times I have breakfast where my blood glucose levels go over 300.’ The app will then pull together different instances and provide insights into the types of breakfast, or the times of the day when you have breakfast, that makes your glucose levels spike.”
After Replica Health users connect their CGM and the insulin pump to the app, it automatically syncs with Apple Health, simplifying the data management for the user. Individuals can then customize their own preferred glucose levels, making the app highly tailored for personalized diabetes management. The platform then pulls in three to six months of historical data about the user’s dietary patterns, giving them a comprehensive view of their health journey.
Replica Health primarily targets individuals with Type 1 diabetes, however, Sam and Brad believe that innovations for Type 1 diabetes will eventually pave the way for advancements applicable to the broader Type 2 diabetes population.
The duo also recognizes the hype and the skepticism surrounding AI, especially among those with chronic illnesses who have been failed so many times by various technological promises. Therefore, Replica Health prioritizes data privacy and security to build trust while aiming to introduce the world to an intelligent, conversational, and data-driven new approach to diabetes management. They envision a future where AI-mediated care becomes mainstream, while also giving patients the final say over their data.
Last Words
Replica Health is more than just a revolutionary app. It represents a transformative idea and a company driven by a strong vision.
The magic of Replica Health is the synergy between Sam’s technical background and his personal experience with diabetes. Sam’s personal experience with the disease, his impressive product and engineering skills, and his forward-thinking approach to machine learning and AI allow the company to envision features and solutions that may initially seem unattainable but ultimately resonate with users.
With its strong emphasis on both the technical and human aspects of diabetes management, Replica Health could be a much-needed step forward in AI-based healthcare.
Join us in welcoming Replica Health to the Type 1 Diabetes Moonshot community.
→ Connect with Replica Health via email
Call for T1D Innovation
Are you a scientist or innovator focused on T1D innovation who would benefit from education about how to navigate and build a company that will be successful in attracting mission-aligned capital, customers, and collaborators to pursue scientific discoveries in the field of Type 1 diabetes? Learn more and apply for a T1D Fellowship.
Become a Health Moonshot Champion
Health moonshots are fueled by passionate families, foundations, and industry organizations committed to achieving health moonshots. Learn how you can join other champions of the T1D Moonshot or one of our other global health moonshots.
Funders: Learn how you can become a Health Moonshot Champion and invest in Health Transformers.
Founders: Don’t make the journey alone. Learn how Health Transformer University fuels your health moonshot.
Follow us on social media for daily updates on Health Transformers: Twitter, LinkedIn, Facebook, and Instagram.
Published: Aug 30, 2023