She had a variety of persistent symptoms—exhaustion, blurred vision, excessive thirst—that couldn’t be attributed to the summer heat. When Oracle Cloud engineer Todd Sharp took his active 13-year-old daughter to the hospital a month ago to find out what was wrong, the diagnosis wasn’t good: She has Type 1 diabetes, which means a lifetime of injecting insulin lies ahead of her.
Being the engineer and avowed tinkerer that he is, Sharp sprang into action, researching diabetes management technology on his laptop while sitting at his daughter’s hospital bedside. Just a few weeks later, he built a smartphone app, called Insulin Helper, to calculate his daughter’s carbohydrate consumption and insulin doses using image recognition, a wireless food scale, and nutritional data. Eventually, Sharp hopes to augment the app with machine learning so that it will train itself, based on his daughter’s personal data, to not only calculate insulin doses, but also predict them.
Oracle Cloud Developer Evangelist Todd Sharp
Oracle“We entered the hospital on Wednesday night, and we were discharged on Friday. By Friday afternoon, I was already formulating the architecture in my brain,” says Sharp, who works as a developer advocate for Oracle. “It’s where my mind goes immediately. How can I use the skills that I have to make my daughter’s life easier?”
Lay of the Land
Every diabetic lives by a unique set of numbers that determine healthy blood sugar levels. These numbers may vary throughout the day, depending on factors such as exercise, food consumed, and insulin intake.
A variety of companies, including Medtronic, Abbot Laboratories, DexCom, and Insulet, already make continuous glucose monitors, insulin pumps, and related diabetes-management devices in a market that’s growing rapidly. But Sharp, the always-curious software developer, wasn’t satisfied with that status quo.
Using the Insulin Helper app on a smartphone or laptop, his daughter first inputs her unique formula for regular insulin dosages throughout the day. When it’s time to eat, she uses the app to snap a photo of her food and weigh it. After its image-recognition services identify the food, the app retrieves USDA nutritional data and calculates the grams to be consumed via a wireless food scale. Then his daughter submits her current blood-sugar level via a standalone glucose monitor. Finally, she gets a recommendation of the appropriate insulin dose based on the number of carbs she’ll consume.
The app links a variety of popular developer tools with powerful cloud services and open data. For example, Sharp used the Node.js Javascript framework to write the glucose formula calculator, along with an assortment of services such as Simple Oracle Data Access (SODA, for retrieving JSON data), Oracle REST Data Services (ORDS), Micronaut Data and Helidon (for coding microservices), Oracle Functions (for serverless calls), and the Oracle Autonomous Database (which stores all of the data).
“The interesting part of this multimodel approach is you can get your data in and out of Oracle Autonomous Database with distributed microservices in any format you need—relational, JSON, XML, graph,” Sharp says. “What you can do with all of this data in three different schemas is combine them, query them, do reports on the data, train machine learning on it—but you’re still in Oracle Database. You don’t have to do events sourcing or some of the traditional approaches that you’d have to do to combine that data in a unified view.”
Sharp is particularly excited about using Micronaut Data, a brand-new, lightning-fast, Java-based microservice that precomputes database queries—essentially, adding annotations to your code that determine which data you’re looking for. “It gets data in and out of Autonomous Database without SQL queries,” he says. “It’s all ahead-of-time-compiled queries based on your schema.”
It took the engineer, who lives on a Georgia farm he has partially automated using the cloud and small hobby CPUs such as Raspberry Pis, two weeks to build the app, followed by a week to refine and test it. “She was diagnosed on July 10. Micronaut Data was released on the 18th, and by end of the 19th I had integrated it into the nutrition service,” Sharp says. “By the 26th I had a working prototype. Since then, I’ve been enhancing, doing bug fixes, refactoring, and improving the user interface.”
Blazing a Trail for Commercial Solutions
Sharp’s initial research revealed that, while technology has evolved to manage the disease, there are missing pieces—and he has the skillset to fill them in. He plans to demonstrate Insulin Helper in a keynote at Oracle Code One, to be held at San Francisco’s Moscone Center on September 17. Sharp says he’s looking forward to getting feedback from any diabetics in the audience.
Open source software has played an important role in diabetes management. While continuous glucose monitors eliminate manual finger pricks and insulin pumps can provide basal dosages (generally, a set amount of insulin delivered hourly) and bolus doses (generally premeal based on estimated carb intake), commercial artificial pancreas solutions (APS) like the MiniMed 670G were preceded by DIY open source solutions such as OpenAPS and Loop/LoopKit. Sharp says he would like to open source his solution eventually, even if just to show other developers how he built it.
But his development work illustrates something bigger: the untapped potential for citizen scientists to solve all sorts of data-driven problems. And who knows—his proof of concept might inspire other FDA-approved solutions. What’s more, the process of writing the app has given Sharp insights into the capabilities of existing commercial products his daughter might use.
Ultimately, Sharp hopes developers and tinkerers won’t forget that there are plenty of real problems awaiting solutions, and that tackling even a small piece of a problem can provide far more motivation than a paycheck.
“I haven’t invented something groundbreaking. I’ve taken things that exist and pieced them together to make a specific solution,” Sharp says. “There’s going to be people in the crowd who understand the disease more than I do. I’ve been dealing with this for a month. They’ve been dealing with it for more than 20 years.”
