In an era where retail transactions exceed $1 billion in volume, senior software engineer and architect Arundhati Kumar shares insights from over 15 years of building high-performance distributed systems. Her journey, from modernizing Fannie Mae’s mortgage platforms to leading global retail technology, reveals how AI, cloud, and self-healing infrastructure are redefining the shopping experience.
From Finance to Retail: A Leap in Scale
Kumar’s transition from banking to retail meant moving from secure, precise transactions to a flood of real-time data from thousands of physical stores. Today, she leads cross-functional teams designing omnichannel point-of-sale applications for one of the world’s largest retailers. The key challenge: ensuring software runs fast and reliably across legacy registers and modern mobile apps, even with unpredictable network connections.
Checkout Doctor: Bridging the Visibility Gap
Inspired by register glitches that stalled checkout lines, Kumar created Checkout Doctor—an automated diagnostic assistant that analyzes local device logs in real time. When a scale or peripheral fails, it instantly alerts associates via mobile devices, shifting support from reactive to proactive. This self-healing approach keeps lines moving and reduces downtime.
AI in Site Reliability Engineering
Large language models (LLMs) are transforming incident management by providing plain-English summaries from error logs, cutting down time to diagnose outages. Kumar emphasizes that AI delivers the most measurable value in noise reduction and early triage—correlating telemetry spikes to pinpoint root causes before thresholds are breached.
Cloud Migration Lessons
Leading a migration from legacy SQL to Azure Cosmos DB, Kumar learned that you can’t copy-paste old database mindsets. Rethinking data organization around partition keys like store numbers and lanes, and accepting eventual consistency, reduced costs and improved availability during peak shopping seasons.
Event-Driven Architecture and Loss Prevention
Event-driven architecture replaces batch processing with instant event streams via websocket servers. This enables real-time visibility and automated decisions—such as pausing a checkout when a barcode mismatch is detected, and prompting associates to intervene. Such techniques help combat the multi-billion-dollar challenge of retail shrink.
The Future: Autonomous Infrastructure
Kumar is most excited about predictive workflow integrity, where machine learning models on edge devices autonomously fix anomalies—like routing around a slowing database—before humans even see an alert. Her advice to leaders: build systems that survive locally without cloud dependency, and treat AI as a co-pilot, not a replacement for human intuition.
Leave a Reply