Shubhankar Shilpi
Zero-Downtime Payments (DR & Rotation), AIOps for High-Value Payments
Abstract:
Every business day, mission-critical payment rails move trillions of dollars; even minutes of downtime can ripple into missed cutoffs, exceptions, and client impact. This keynote presents a production-tested blueprint for keeping high-value payment systems “always on” while improving detection and recovery with AIOps. Drawing on programs across ACH (PEP+), ACI Online Treasury Manager, and SWIFT/Fedwire, I show how to make disaster recovery (DR) and site rotation routine, measurable practices—not emergency rituals. The approach combines SLO-driven design, deterministic controls (idempotency, replay, message ordering), and dual-run/canary cutovers with automated reconciliation to guarantee correctness under load. On the operations side, we fuse heterogeneous telemetry (CA-7, JES2, SMF, MQ, SWIFT, API traces) into a canonical stream and apply rules plus explainable ML to surface alert-as-story incidents that operators trust. This includes lessons from my patented Batch Monitoring Tool, which materially reduces mean-time-to-detect and recover across thousands of nightly jobs. Attendees leave with a pragmatic checklist: setting journey-level SLOs; practicing zero-downtime DR/site rotation; designing for throttling, backpressure, and fairness; normalizing and correlating signals for faster root-cause; and governing AI with drift monitoring, challenger thresholds, and human-in-the-loop approvals—so real-time payment experiences are delivered safely at scale.
Profile:
Shubhankar Shilpi is a Software Engineering Director II at Truist (USA) with 16+ years in enterprise payments, site reliability engineering, and AIOps. He has led large-scale modernization and resilience programs across ACH (Fiserv PEP+), ACI Online Treasury Manager, and SWIFT/Fedwire, delivering zero-downtime disaster recovery and routine site rotation, and converting heterogeneous telemetry (CA-7/JES2/SMF/MQ/SWIFT/API) into explainable, actionable incidents. Shubhankar is the sole inventor of the Batch Monitoring Tool (USPTO App. 18/488,784), a framework that significantly reduced mean-time-to-detect and recover across 3,500+ nightly jobs, protecting high-value transactions and improving SLA adherence. His work focuses on trusted intelligence: pairing deterministic controls with ML under model-risk governance, drift monitoring, and human-in-the-loop approvals, while emitting audit evidence via policy-as-code. A frequent speaker, reviewer, and panelist, he translates legacy-to-AI transformations into field-tested playbooks leaders can apply immediately. His current interests include federated learning for operations, graph-based correlation for real-time systems, and incident replay analytics.