The Fifth Elephant 2024 Annual Conference (12th &13th July)

In today’s data-centric world, businesses rely on personalization now more than ever. Whether it’s personalizing user experiences, optimizing operations, or predicting market trends, data plays a pivotal role. To harness the full potential of data, organizations are turning to real-time feature stores. In this talk, we’ll explore what real-time feature stores are, why they matter, and how we at Zupee have leveraged real-time feature stores for various use cases.

Zupee is a rapidly growing real money gaming platform that caters to a growing user base of more than 100 million registered users. We have been building data powered models to enhance the overall user experience and improve the integrity solutions on our app. These include multiple ML models to improve user recommendations, system integrity solutions etc. The full potential of these models can be leveraged when deployed on real-time data. Serving features on a reliable, real-time and centralized basis would be necessary for these to work. This calls for a Real Time Feature store to be built.

Outline

• The audience for this talk would be Data Engineers, Data Scientists or ML Engineers who are building platforms to provide personalized experiences in real time
• To Build a reliable and scalable real time data infrastructure that powers ML use cases is incredibly challenging to build and this talk would be helpful to folks who are thinking of building such user experiences
• Providing guidance on high level design and talking though the challenges in execution so that folks can avoid the common pitfalls while building real time feature stores

Impact

By leveraging the real-time feature store, we are able to activate multiple ML use cases such as fraud detection, player protection, personalized experiences, recommender engines etc. on a real-time basis and also serve these centralized features for other use cases such as:

1. Enhanced Decision-Making
2. Efficient Model Training
3. Ensuring Consistency, Reliability and Reproducibility
4. Enabling Governance