Scaling AI for High-Stakes Real-Time Payments

Learn how to implement AI systems in high-stakes, real-time payment processing environments through this 25-minute conference talk from StaffPlus New York 2025. Discover the engineering challenges and solutions involved in scaling machine learning for fraud detection within Stripe's critical "charge path" infrastructure, which processes over $1.4 trillion in annual transaction volume. Explore the complexities of working within a shared Ruby monolith that serves as the backbone for nearly every transaction on the platform, where even minor changes require careful coordination across multiple engineering teams and hundreds of developers. Understand how to navigate legacy architectural constraints and Ruby's concurrency limitations while maintaining strict latency requirements for business-critical systems. Examine the collaborative approach used to align cross-functional teams through event storming workshops, uncovering tribal knowledge and identifying architectural leverage points for safe parallelism implementation. Gain insights into building multi-quarter roadmaps that coordinate efforts across teams, minimize conflicts, and accelerate the delivery of AI-powered fraud prevention improvements. Master the strategic and technical considerations necessary for delivering machine learning innovations within latency-sensitive, high-volume payment systems while balancing performance constraints with business impact.