Return To The Wire

YHack Winner

When first year Solomon Greenberg and his friend Benjamin Ward set out to solve one of the biggest hassles facing diabetics, they worried they were being far too ambitious. And it didn't help that they'd have to achieve their goal in 36 hours.

Greenberg and Ward went to work on Friday, November 30, at the YHack hackathon at Yale University. When they were done on Sunday morning, they'd not only achieved success, they'd nabbed YHack's Grand Prize, beating out some 1,400 competitors. They also won the Best IoT Hack Using a Qualcomm Device and Code4Good Civic Track Original Project awards. "Somehow we figured out how to make everything come together and get something working,” Greenberg said.

Greenberg said he got by on four hours of sleep, a gallon of Arizona Iced Tea and spicy Doritos. He said he found the experience thrilling. "For 36 hours you don't worry about anything else except what you're working on," he said. "There's not many places you can go for pure concentration like that. It was really awesome."

When he was 11, Greenberg was diagnosed with type 1 diabetes, a condition where the body doesn't produce enough insulin and blood sugar, or glucose, levels can rise dangerously high. To prevent this, Greenberg pricks himself several times a day with a needle-shaped sensor that measures his glucose. Based on the results, he injects himself with insulin using a pump he wears on his body.

If Greenberg forgets to check his measurements or miscalculates how much insulin he needs, it can cause serious health problems. "It's a huge stress,” Greenberg said. "It's a big burden in my day-to-day life.”

According to Greenberg, the "holy grail" in diabetes care is automating the process of measuring glucose and injecting insulin. No human intervention would be necessary. This is what he and Ward achieved at a very rudimentary level at the hackathon.

They started by connecting a Dragonboard 410C microcomputer to a glucose monitor. The data was then uploaded to the Google Cloud Platform where it was plugged into a series of algorithms previously devised by medical researchers to simulate the functions of the pancreas, which is where insulin is produced.

Greenberg and Ward's programming featured machine learning, which is vital because the algorithms need to adjust to an individual's unique body chemistry and eating habits.

The computations done in the cloud produced a value for how much insulin the pump should inject into the bloodstream.

As it happens, the insulin pump has a remote control. Originally, Greenberg and Ward planned to download the information from Google's servers to the Dragonboard and then have the Dragonboard communicate with the remote via radio signals.

This turned out to be a nightmare because, Greenberg said, “the remote's communication protocol is really, really complex. There was no way we could reverse engineer it in 36 hours, let alone fix any of the other parts of the project that weren't working.”

Panic and despair set in until Greenberg proposed to Ward that they “crack the remote open and wire it straight to the Dragonboard because that's totally doable right now.” It was a bit crude, but it worked.

Once they'd figured this all out, Greenberg did what he now recognizes was "from a health and safety standpoint, not one of the smartest things I've ever done.” He tried out the apparatus on himself.

For three hours, his blood sugar and insulin levels were monitored and controlled by a computer. But it worked.

Greenberg and Ward, a first-year student at Worcester Polytechnic Institute, plan to continue working on their invention. "I'm absolutely going to continue to develop it,” Greenberg said. "I want it to become something I can use on a day-to-day basis.”