Building a Resilient, Real-Time Fraud System at Block

Implementing effective fraud detection in a complex, evolving ecosystem like Block's is challenging, especially when it comes to real-time analysis and adapting to new fraud patterns. Niccolò Ronchetti, Staff Machine Learning Engineer at Block, presents an innovative approach: In this talk, he introduces a Bayesian-based approach for fraud detection at Block, designed to combat fraud across their diverse platform while maintaining scalability and adaptability.

Niccolò will explain how Bayesian methods provide a probabilistic framework for fraud detection that excels in complex environments. Using Block's multifaceted ecosystem as a case study, he'll demonstrate how this approach allows for real-time risk assessment that evolves with user behaviors and emerging threats. Niccolò will also discuss why Tecton was chosen as the cornerstone technology for implementing this system, highlighting its ability to manage and serve machine learning data at scale.

In this session,  you’ll:

• Understand the advantages of Bayesian approaches in complex fraud detection scenarios.
• Learn how to scale fraud detection systems to handle diverse and evolving ecosystems.
• Discover strategies for implementing and maintaining real-time fraud detection using Tecton's feature platform.
• Gain insights into overcoming technical challenges in large-scale machine learning deployments.

To illustrate the practical application of these concepts, Niccolò will walk through Block's implementation process with Tecton. He'll showcase how Tecton's capabilities enable Block to deliver real-time, adaptive fraud detection across their platform, allowing for rapid response to new fraud patterns while maintaining system reliability.