Outsourcing ML Experts to a U.S. Virtual Solutions Provider for Retail

We partnered with a U.S. start-up, an innovation leader in 3D simulation software and virtual market research for retail, to enhance their data science capabilities.

As the creators of a best-in-class virtual store research platform for capturing shopper behavior, our Client sought to further refine their models and explore innovative approaches to representing store layouts. To support their goals, they outsourced a data scientist from our team, who seamlessly integrated into their data science division. This collaboration resulted in improved model performance and the successful testing of new techniques for optimizing store layout representations.

Our Client faced three critical challenges.

• First, they needed to enhance the quality of their models to deliver more accurate and actionable insights into shopper behavior.
• Second, improving processing efficiency was essential to ensure their platform could handle the complex simulations required for large-scale retail environments.
• Finally, preparing a new solution for on-shelf product arrangement strategies was crucial for their virtual store research platform to maintain their competitive edge in the retail market

To address those challenges, one of our data scientists worked closely with their team for six months, contributing across multiple critical areas. He worked on refining and optimizing machine learning algorithms, ensuring they delivered more accurate insights into shopper behavior.

Additionally, he collaborated with another coworkers to improve the readability and simplicity of the production-ready code.

All solutions were prepared for optimal operation in a cloud environment, enabling scalable and efficient processing for the virtual store research platform.  

• Using enhanced models has significantly improved the accuracy of predictions, providing a much better representation of actual customer behavior when purchasing products. 
• By analyzing and addressing bottlenecks in the code, we were able to streamline the process, significantly speed up prediction generation, and reduce the prediction cost.  
• Introducing a new reinforcement learning model made it possible to identify optimal product arrangement strategies, effectively boosting key sales metrics.
• All solutions have been adapted for deployment in a cloud environment, enabling rapid scalability of the project.