The virtual environment, known as Ahavi, has been in development for the past few years.
“This is an environment that allows our organization to assess the efficacy of AI models against our patient population prior to ever having to deploy it against our actual population,” Jeff Jones, senior vice president of product development at UPMC Enterprises, told Becker’s.
Ahavi leverages secure testing, de-identified patient data and strict governance protocols to validate AI solutions, according to Mr. Jones. The goal is to create a controlled space where AI models can be trained, fine-tuned and assessed before implementation in a real-world healthcare environment.
Given the size and complexity of UPMC, implementing new technologies can take more than a year. However, Mr. Jones emphasized that Ahavi’s controlled environment allows for fine-tuning models, addressing AI governance concerns and mitigating bias before broader deployment.
“We allow these organizations to take the models and the insights derived from our population and bring them to their customer bases, ultimately reducing the research and development effort on behalf of many of these AI-focused health IT organizations,” Mr. Jones said.
Currently in what Mr. Jones refers to as a “soft launch,” Ahavi will debut at the ViVE conference. UPMC Enterprises is actively engaging with multiple organizations in contracts.
“I can’t stress enough, we’re looking for organizations that we’re willing to partner [with], to work with us and to exchange ideas as to how we further enhance these models,” Mr. Jones said. “Our goal is to collaborate in a way that ensures insights can benefit future customers while allowing partners to work with us in properly calibrating the technology before deploying it across the UPMC system.”