Healthcare giant Johnson & Johnson is injecting data science across its business to improve its manufacturing, clinical trial enrollment, forecasting and more.
“I actually like to call it decision science,” said Jim Swanson, the company’s executive vice president and enterprise chief information officer, in a panel discussion at the most recent NVIDIA GPU Technology Conference. “It’s not just about creating a model — it’s actually what decisions and insights you’re trying to derive from these models.”
Machine learning operations, known as MLOps, is a set of best practices for businesses to run AI successfully. With an MLOps strategy, companies can harness data to answer difficult questions and measurably boost business operations.
Johnson & Johnson, for example, has formed an internal Data Science Council and developed something Swanson calls “bilingual data scientists” — roles that combine deep domain expertise with data science skills.
“They have that understanding of the main business problem, and they have the skills to do data science,” he said.
This strategy helps integrate a company’s data science community into the business workflows, enabling faster application of machine learning models, feedback and impact, Swanson said. It also helps overcome hesitation to data science adoption.
“You’ve really got to show them by proving over and over again: Hey, this model doesn’t replace the valuable asset of people skills you have in your business domain, it gives you longitudinal views you can’t get on your own,” he said.
As companies scale up their adoption of MLOps, they need a powerful AI infrastructure to support their engineers and data scientists, said panelist Nick Elprin, CEO of Domino Data Lab.
“So many companies, tragically, waste precious engineering resources trying to build this tooling themselves, and it’s a lot harder,” he said, recommending that companies get started with a third-party platform like the NVIDIA GPU-accelerated Domino Data Lab. “Your engineers are going to be creating much more value when focused on problems that are competitively differentiated and unique to your business.”
To help businesses get started with MLOps, NVIDIA provides a suite of open-source tools on the NGC software hub for managing an AI infrastructure based on NVIDIA DGX systems. Healthcare companies and medical researchers use DGX systems and the NVIDIA Clara application framework to run healthcare models that parse electronic medical records, boost computational drug discovery and power AI-enabled medical devices and imaging workflows.
From Vision to Business Impact
In addition to embedding over 1,000 data scientists in its business, Johnson & Johnson is working to build digital literacy across the whole company — helping employees understand the potential of machine learning models in action.
Swanson gives the example of an internal hackathon Johnson & Johnson held to improve forecasting in its vision care business. By better predicting how many customers will need each product in its line of Acuvue contact lenses, the company can more efficiently manufacture the ones people want.
For every percentage point Johnson & Johnson improves forecast accuracy, the company boosts revenue, “because you have the right product going into the market,” said Swanson.
Dozens of teams across the company — most outside the vision care business — signed up for the hackathon, which used Domino Data Lab’s Enterprise MLOps platform. The procurement team came up with the best model.
“We solved a really big problem with real impact, and they learned some new tooling that they hadn’t known before,” Swanson said. “With a simple project aligned to a really significant outcome, you can do amazing things.”