alt

Discover the ML Platform That’s Helping Walmart Take Over Global Retail
 

With over 11,500 stores in 27 countries, Walmart enjoys a sizable retail market share. Powering this success, is a nexus of technology and innovation that has helped us to scale operations and create a globally influential brand. But in an industry with unprecedented access to data and analytical power, staying relevant means evolving every aspect of the organization - from operational efficiency to user experience.
 

At Walmart, a key part of this evolution involves using Machine Learning (ML) and algorithmic data analysis to build immersive and relevant experiences for both customers and associates.
 

With ML emerging as a prime driver of innovation in retail, the importance of having a powerful collaborative tool for ML developers is the need of the hour.
 

ELEMENT was built as an intelligent platform to democratize access to all AI and ML initiatives within the organization.
 

The evolution of Element
 

Element, a collaborative, end-to-end ML platform, gives engineers access to vast library modules, application frameworks, and tools that help rapid development and deployment of new solutions. Faster development, in turn, enables faster decision-making, spawning a virtuous cycle that is at the core of sustained, high-performance innovation.
 

Initially the Element team comprised of just 15-16 engineers running beta-tests. After going live in 2018, the platform has rapidly evolved.  Element was conceived in line with Walmart’s philosophy of building products, as opposed to buying them, and delivers significant value to both business and technical users, including:

  • Scalability: The platform was initially built with limited technologies to support limited functionalities. Since then, new features and centralized resource allocation capabilities have been added to scale the platform’s performance alongside expanding business needs.
  • Security: Element offers fewer security risks than third-party platforms, as it’s an in-house engineered product.  On a related note, reusable data connectors to sources such as Data Lake, Teradata, Hadoop clusters, and DB2, enable secure integration with enterprise data systems.
  • Ad hoc Enhancements: In order to support crucial business processes and ad hoc development requirements, new features are quickly rolled out and deployed to Element. In fact, on average a new feature is added every 2 weeks!
  • Stability: Element’s response team quickly tackles any service delivery bottlenecks, while complex code releases are staggered, ensuring minimal downtime.

So, how does Element help solve real-world business problems?
 

In the absence of a common platform, the sheer volume of data from different business units made it difficult for Walmart data scientists to engage in meaningful collaborations. Element brings together data and various AI and ML technologies. It creates an integrated development environment where disparate teams can work together and address practical problems, including:

  • Aisle assortment: Prior to the launch of Element, the  team managed aisle assortment on the basis of intuition and experience. However, rising revenue goals and customer expectations made analytics critical to achieving ideal product placement. Via Element, the team was able to correlate hundreds of variables, including store locality, season, new product launches, past sales trends, and customer behavior, to optimize aisles for each store.
  • Wait-time at stores: Teams always want to reduce order pick up times, especially for customers who opted to order online and pick up in-store. By adopting ML technologies to analyze the peak order times against store associate bandwidth and other key experience enablers, customer wait-times have significantly reduced.
  • Personalization: Teams have always wanted to extensively personalize the shopping experience but were hamstrung by a lack of insightful data. As Element grows more and more intelligent, with data inputs from multiple verticals, it better understands each customer, alongside their preferences and their needs. These insights help Walmart cross-sell and upsell products, driving customer satisfaction and improving revenue.

Why do Walmart engineers love working with Element?
 

It’s hard to find a developer at Walmart who doesn’t enjoy working with the platform. From analyzing and segregating data to developing and deploying models, Element supports every team working on AI and ML initiatives with functions that encompass:

  • On-demand computation of resources: This machine learning platform uses cloud technology to simplify resource allocation and access management. Scalable bandwidth offers a great deal of workspace flexibility, while eliminating dependencies on-premise infrastructure
  • Integrated APIs: The platform allows quick sharing of project artifacts, including data sets, workflows, and notes. Element’s API-based integration with enterprise systems also makes it easy for different teams to access data from across the enterprise, on-demand.
  • Smarter workflows: Element provides an integrated development environment (IDE) which makes it easy to create complex training and scoring workflows from the user interface.
  • Easy model management: The in-built version control tool makes it easy to experiment with new models while capturing metadata for further analysis and eliminating data duplication.
  • Production deployment: The platform supports different deployment types (batch scheduling, real-time, R-shiny, and Python) and also uses APIs to trigger workflows from external applications.

Element, the motherboard of all AI and ML initiatives at Walmart, converges talent and ideas from all over Walmart on to a single platform. This helps our analysts and developers quickly build solutions to some of the most critical enterprise issues - from code reusability to optimized infrastructure utilization. Element also helps standardize processes, offers a digital sandbox for production conceptualization, and opens new doors of collaboration among data scientists. It’s a living machine that grows and evolves with Walmart as we continue to redefine the face of global retail. 

alt
Rakshith UJ 

Sr. Product Manager – Platforms