MLOps Key to Creating AI Innovation Opportunities, says Northwestern Mutual’s Anju Gupta

For the world of AI and model building, the famous motto, “Move fast and break things” by Facebook founder Mark Zuckerberg is a sure path to destruction. Stakes are now higher than ever for AI, and model-driven businesses must be careful to test and confirm the safety and fairness of technology touching nearly everyone’s life. How can stakeholders do this while simultaneously pressing business initiatives as quickly as possible? We learn how in Domino’s new Financial Services and Insurance edition of the Data Science Innovator’s Playbook from Anju Gupta, Vice President of Data Science and Analytics at Northwestern Mutual.

Meet Anju Gupta at Northwestern Mutual

Anju Gupta is a data science innovator at Northwestern Mutual, a Fortune 100 leader named as one of the World’s Most Admired Companies for seven consecutive years. The company’s enterprise data strategy is to drive business outcomes by leveraging data, predictive analytics, quantitative research, and a customer-centric lens.

As detailed in the Data Science Innovator’s Playbook, Gupta says the company’s innovation includes accelerating use of AI technology by integrating data science throughout long-standing organizations using MLOps, explainable AI, and a framework for organizational practices. This program has streamlined approval and acceptance of machine learning and artificial intelligence models both inside and outside the businesses where Gupta has worked.

“With MLOps, we can innovate much more quickly now and work on problems that we couldn’t even think of solving before,” Gupta says.

Using MLOps to Drive the Pace of Model Innovation

MLOps accelerates the innovation rate for Northwestern Mutual, Gupta says. Before MLOps, the company’s data scientists were limited to sequential, manual processes of building and deploying models. Now, with the data science platforms, teams have far more accessibility to leverage each other’s work instead of reinventing the work. “With the newer tools and platforms, it’s much easier to collaborate and iterate together,” says Gupta. “This has unlocked the potential for us in how we collaborate with our business partners as well.”

As an innovation change agent, MLOps has served a pivotal role, from data ingestion to building models and streamlining, optimizing, deploying, and performance monitoring. Gupta says MLOps frees up a lot of time for data scientists to look at other business problems and provide opportunity and white space for R&D by avoiding repetitive tasks such as reserving infrastructure or configuring the environment for each project.

“With MLOps, we can innovate much more quickly now and work on problems that we couldn’t even think of solving before.”

Download our free eBook to read more details of insights by Northwestern Mutual’s Gupta and other innovators, advisors, and industry experts at the top of their game in data science within the financial services and insurance industries.

Having a foundational MLOps platform is essential for Northwestern Mutual’s model lifecycle. In addition to this core technology engine, the company’s strategy required creation of a framework for deploying models using APIs, which the business uses to make decisions. This approach can be useful to any organization looking for ways to reduce model risks – especially those driving potentially unpleasant oversight questions that may appear in the news.

“We have a really good model risk framework that has set us up for success,” Gupta says. The framework includes an executive review committee to ensure that all models are managed not just for utility, but for conformance with other business requirements such as security and compliance. Some major risk factors covered by the framework include managing:

• Where data is stored
• Features used
• History of the model
• Inventory in the MLOps framework
• Sensitivity factors
• Bias testing

Conclusion

Gupta says data science-driven innovation at Northwestern Mutual is helping the company better serve its customers and other owners. One example is tapping the potential of natural language processing. It entails training a model to machine read oft-handwritten Attending Physician Statements, which automates their translation into Electronic Health Records to help quantify risk classes or process claims for more efficiency. We invite you read Gupta’s full story along with profiles of other innovators; download our free eBook. Their stories will provide you with many ideas on how to move fast and not break things by accelerating innovation with MLOps in the Financial Services and Insurance industries.