Earlier this month, Dave Esra's startup, BobiHealth, competed for $20,000 as part of Geekdom's pre-accelerator program. He was one of six startup founders who offered polished, five-minute pitches.

BobiHealth did not win.

After the event, as Esra was thanking advisers and chatting with fellow founders, a person tapped him on the shoulder and said he wanted to invest in the company.

“And so that's the main point, right?” Esra said later. “Getting on investors' radar.”

The mission of San Antonio-based BobiHealth is to make pregnancy safer. Its app collects vital signs and other data about a pregnancy, then uses machine learning and AI to predict the risks of adverse events, like preeclampsia, infection, mental health concerns, blood clots and hemorrhaging.

The BobiHealth app collects vital signs and other data from the pregnant person, then uses machine learning and AI to predict the risks to the user of adverse events, like preeclampsia, infection, mental health concerns, blood clots and hemorrhaging.
The BobiHealth app collects vital signs and other data from the user, then uses machine learning and AI to predict the risks of adverse events, like preeclampsia, infection, mental health concerns, blood clots and hemorrhaging. Credit: Courtesy / BobiHealth

Unchecked, those conditions too often lead to death. The World Health Organization found that in 2020, almost 800 women died from preventable causes related to pregnancy and childbirth — every single day.

“And 94% of those are preventable deaths,” Esra says. The numbers are slightly better in the U.S., he noted, but this country has the worst maternal mortality rates of all industrialized countries.

High-risk pregnancies

Esra's first experience with a high-risk pregnancy was his wife Sarah's. She struggled with several complications, including gestational diabetes. Today, their son Cooper is a healthy 12-year-old, but the stressful and isolating experience stuck with Esra.

During the pandemic, he was working with another startup, trying to study the effects of the Covid-19 vaccine on pregnancy. It was there, he said, that he began to truly understand the “huge gap in knowledge we have about maternal health — and women's health in general.”

Later, doing post-graduate work at UT Austin, Esra and his fellow students were challenged to solve a problem with artificial intelligence. By then, he said, “it was obvious to me what problem we needed to solve.”

Sarah Webb-Esra holds the ultrasound photos from her pregnancy with her son Cooper Esra. During this pregnancy she developed health complications, including gestational diabetes. Today her son is a health 12-year-old boy.
Sarah Webb-Esra holds the ultrasound photos from her pregnancy. During this pregnancy she developed health complications, including gestational diabetes. Today her son is a healthy 12-year-old. Credit: Bria Woods / ISF FORUM

He realized that with smart phones, which 95% of Americans possess, “we could meet people where they already were.”

Esra eventually submitted a patent for a smartphone app that collects data and uses machine learning to predict adverse events, and incorporated BobiHealth — pronounced Boh-bee — in May of 2023.

Earlier this month, the BobiHealth app launched in the Apple Store, available in the U.S., India and the Philippines. The app is free to download. The user shares data either manually or through an Apple Watch, Fitbit or other wearable health device.

The app has now been downloaded close to 500 times, Esra said.

Addressing privacy concerns

Esra is keenly aware of the privacy concerns that come with entering your personal data into an app, and says the company's unique machine learning architecture is what differentiates it from other data-gathering tools.

The key, he said, is an architecture he called “federated machine learning systems with differential privacy.”

Susanna Disesdi Cox is an artificial intelligence architect and security researcher based in Chattanooga who served as BobiHealth's data pipeline architect.

She explained that with federated machine learning, instead of taking all the data into a central model, “mini-models” are instead sent to each user's phone. Their data trains the model on the phone, and then only the model is transmitted back. It's then aggregated with all the other mini-models to enable the predictive analytics that alert the user if they need to, for example, keep an eye on a vital sign, call their doctor or head straight to the emergency room.

An example of a mini-model, Esra said, is the auto-correct on your phone. The recommendation it gives the user is based on their own past typing, not everyone's.

That all means a user's data never leaves their device, she said. Adding differential privacy — a mathematical way to protect individuals' data when it's being used in large data sets — further ensures privacy.

While federated machine learning and differential privacy are already being used in other cases, she said, BobiHealth developed “a security back-end that would additionally protect data anywhere there are edge devices or sensors that are gathering data.”

BobiHealth has applied for a patent for the technology, which Cox called “defense-grade;” Esra said they plan to pitch it to the Defense Department.

Social impact or big bucks?

The company's privacy architecture isn't the only technology that could have uses outside the app, its supporters say.

David D'Annunzio, a corporate executive who volunteers as a business mentor through the U.S. Small Business Administration's SCORE program, has been mentoring Esra for about a year and a half.

Dave Esra presents the BobiHealth app during a pitch competition in May as part of Geekdom’s pre-accelerator program.
Dave Esra presents the BobiHealth app during a pitch competition in May as part of Geekdom’s pre-accelerator program. Credit: Bria Woods / ISF FORUM

“The real secret sauce of what he's doing there is the technology they're developing, which will have multiple applications beyond pregnancy,” he said. “That's the golden nugget.”

Dr. Ken Marriott, an emergency medicine physician who serves as the company's medical director, said something similar.

“I think the underlying model ... could change the way we look at a lot of disease processes,” he said. “If we start actually getting more data, we might pick up on subtle things, like changes in vital signs or in symptoms over a longer period of time. And that might provide the warning that people need to have better outcomes.”

The company is already looking at ways the technology could help caregivers of those with autism, Esra said: “It really is plug-and-play.”

Esra said that as he's participated in various pitch events around Texas he's shifted his message.

His early pitches leaned hard on the social impact he hopes BobiHealth will make, as a low-cost way to help populations that suffer the highest maternal mortality rates, such as in rural areas and among people of color. It is this opportunity that drives him, he says.

But many investors want to hear about the money-making opportunity of health data collection and BobiHealth's proprietary security architecture, he said.

“It's definitely been a learning experience.”

This story has been updated to correct the spelling of Susanna Disesdi Cox's name.

Tracy Idell Hamilton covers business, labor and the economy for the ISF FORUM.