��������

Applications running on artificial intelligence can make unexpected shifts in performance, which can lead to inaccurate results for users and create headaches for their creators.  

Mike McCourt (AMAT ’07) helped to found the AI enterprise testing platform company and serves as its chief technical officer. In his role, he applies his mathematical knowledge and skills to optimize these apps and help clients ensure their products are performing in the ways that they want.    

“Let’s say you’ve built a GenAI platform, you’re happy with it, and you’re ready to launch. But how do you know that it is going to do what you want it to do?” McCourt asks, referring to generative AI platforms that provide developers with powerful optimization tools. “You’ve got to be thinking about how to know the system is working the way that you want it to.”

Using mathematical tools and examining key statistical data, Distributional can test whether a client’s application is going to produce outputs that users will find helpful. Distributional’s other mathematical tools also monitor how the applications perform once they’ve been launched to ensure that they continue to meet users’ needs.

The rise of generative AI has made it easy for people with little computing knowledge to use it and rely on it to get questions answered. McCourt says that the ability for the technology to recognize natural language opens it to many more users.

“Large language models using natural language are powered by a lot of math,” McCourt says.  

And that simple fact has earned him a lot of professional acquaintances.    

Having advanced skills and knowledge in mathematical concepts, McCourt has been able to collaborate with specialists in fields such as materials science, health care, team management, and electrical engineering throughout his career.

“I didn’t have a background in these fields, but I would read journals and learn about these fields,” he says. “I was very lucky to meet and work with ambitious people who may have no background in math, and together we could move the field forward.”

Now, he is applying his skills in AI as increases in demand, and increases in performance expectations, force these AI systems to produce more reliable and more accurate results.  

But keeping up with the rapid gains in performance and expectations is a challenge, McCourt says.

“Just a decade ago, we couldn’t feed a computer an image and ask it for a caption,” he says. “Now we can give it a caption and get an image.”

The trajectory of the technology shows promises in solving big problems, such as developing new, more resilient antibiotics. But McCourt says he is also excited about the technology’s ability to solve small problems that can make everyday life a little easier.

“Imagine an Alexa-type device that can tell when you’re upset and starts playing calming music or can tell that you have a headache and automatically dims the lights,” he says. “Everyday life can be exciting, too.” —Casey Moffitt