Home
>
Nitin Kumar, Vice President, AI and Data Engineering, Visa Inc.

Nitin Kumar, Vice President, AI and Data Engineering, Visa Inc.

Describe your career to date

I have been fortunate to always work in roles that relate to using data for decision making. I started in personal finance product management portfolio management and customer graduation. I was proud to work in a business line that helped create and support many small businesses. The responsibility for borrowers using credit appropriately also lies with the lender, and this started my interest in using data to make decisions whether at an individual level or to inform strategy.

I spent time in management consulting advising large companies in customer relationship management and sales compensation. I was fascinated by the number of examples of lack of outcomes that could be traced back to data quality, systems, and artificial intelligence (AI) and machine learning.

I joined Visa in their advisory business focused on building data science and data engineering teams to support clients. In this role, I focused on leveraging product, data science, data engineering and data visualization teams to drive outcomes for our clients. Many of these are a result of business strategy driven by AI and machine learning models at global scale. I also focus on my aspects of enterprise data strategy, like data mesh.

How are you developing the data literacy of your organization, including the skills of your data teams and of your business stakeholders?

We have a lot of data, not just generated as part of transactions, but also other types of customer interactions like usage of benefits. Our teams have evolved significantly over the years in leveraging this data for the benefit of our customers and to make better decisions for our business. For instance, our AI and machine learning models in the risk domain support banks in identifying fraudulent transactions. The same kind of models and analyses also help us to build products that discourage fraud.

Data literacy is a team effort. We have several programs underway to sharply define the skills and talents for our data practitioners; they are not all data people. We start with having many job families with definite skills like data product managers, data engineers, data scientists, researchers, MLOps specialists, generative AI (genAI) specialists, data visualization specialists and more. Finer separation in roles enables more targeted training, development, role profiles and leads to better overall effectiveness of teams.

Our stakeholders support us. With them, we focus on the agile partnership model where joint teams for business outcomes have data practitioners embedded. (as opposed to separate data teams). This enables a deeper integration of teams with business outcomes. We have learnt that overtime this proliferates a data-first thinking into product design.

Have you been able to fix the data foundations of your organization, particularly with regard to data quality?

Data foundation is a journey, there likely is not a finite step when we will solve it. We constantly deal with changes in the data environment. For instance, we build new products or acquire companies frequently. Both these scenarios require us to think about these new data streams that will be related to existing data, infra and applications. The end goal is the same. To have quality data available for any business processes and AI and machine learning models.

One key aspect of this strategy is our work on data mesh. We are building a new platform for data consumers to ingest, process, govern and monitor data for products and solutions. By following a platform approach, the data office aims to be the enabler for others. Users of this new platform can self-serve and consequently avoid the need to scale up and down central teams with the volume of data related activities.

The other major advantage of this approach is that we can centralize certain features like data quality monitoring against a quality framework. This has enabled significant value in certifying certain data sources and proactive monitoring of data feeds by the users themselves. We believe this partially decentralized data and quality approach will pay many dividends in application-specific data quality, especially as the standards of data quality differ significantly by use.

Nitin Kumar
has been included in:

100 Brands 2024 (USA)