Josh Poduska has been busy.
In the three years since he joined Domino Data Lab, he’s watched the volume of successful machine learning projects among the company’s customer base grow in scale. Domino Data Lab’s platform has enabled clients to apply machine learning in a bevy of new use cases. Think: building an intricate, AI-based research model for cancer or teaching an AI system to fly a military jet.
But Poduska said those projects, though impressive, aren’t necessarily the most surprising part of how the company's customers are applying machine learning.
“What is really surprising to me is the volume of smaller machine learning wins that have incremental value,” Poduska said. “No small project on its own is surprising, but the volume of successful projects is. Companies are figuring out how to scale machine learning, and we are helping them get there faster than their competition.”
Built In SF connected with Poduska to learn more about Domino Data Lab, machine learning — and the unexpected ways the technology can be integrated into their clients’ work.
Domino Data Lab is an enterprise data management platform built for data scientists in every industry (more than 20 percent of Fortune 100 companies use Domino). Poduska continues to bring significant value to his customers by maintaining the overall system health of machine learning models so that when problems arise, their customers are ready.
What’s a surprising or interesting way your company is using machine learning?
We have several projects in our customer base that fit what people would typically think of as a surprising use of machine learning. A complex AI-based cancer research model and an AI systems that has learned to fly military jets. What is really surprising to me is the volume of smaller machine learning wins that have incremental value. This flies in the face of what is often reported in the press as the systemic failure of getting machine learning projects in production. The cumulative benefit of these small wins is more than that of the fewer home-run projects.
One customer has a portfolio of 25 customer service brands to manage and sees intense demand for machine learning. Examples include powering forecasting tools on websites and increasing product search relevancy. Another customer, a global investment management and insurance company, has seen multiple wins in the form of automating the tone of customer communications and recommending resumes to review when hiring. No small project on its own is surprising, but the volume of successful projects is. Companies are figuring out how to scale machine learning and we are helping them get there faster than their competition.
What impact has machine learning had on your business, product or the customer experience?
One nice thing about these smaller projects is that they often lend themselves to A/B testing so their impact on the business can be easier to measure than large, complicated projects. We worked with a manufacturing contractor to implement machine learning solutions across supply chain risk forecasting, machine failure prevention and other projects. A formal, joint business value assessment found $20 million in annual benefit to the contractor with an 8x return on investment.
Companies are figuring out how to scale machine learning and we are helping them get there faster than their competition.”
What’s a part of your tech stack that you really enjoy working with, and how are you applying it in your work?
I really enjoy working on the validation and maintenance of machine learning models within our tech stack. I feel that this part of our platform brings significant value to our customers. As organizations implement more and more models, validating them for usability, effectiveness, ethics, and bias becomes very important. So does the need to maintain these models in production. Checking the overall system health of models and knowing what to do when things go wrong are tough problems to solve but exciting ones that we are working on at Domino.