Drove a complete digital transformation of Hertz's global rate engine, then engineered a 6,000% throughput improvement — achieving >3,000 reads/sec and >2,500 writes/sec with p95 latency under 30ms.
The Problem
Hertz's legacy rate engine was collapsing under load at ~50 requests per second. The system needed to handle worldwide localized pricing data with real-time consistency across 10,000+ locations, processing millions of rate queries daily. During peak periods (holidays, weekends), the system degraded severely — causing lost revenue, abandoned bookings, and customer frustration. The entire technology stack needed modernization: aging systems, brittle integrations, and no path to scale.
The Solution
Led a multi-phase digital transformation: first upgrading and stabilizing legacy systems, then re-architecting for scale. Designed a distributed rate engine utilizing eventual consistency to handle >2,500 pricing writes per second while maintaining real-time accuracy for rate shopping. Engineered a multi-tier caching strategy using Redis for hot paths and Cloudant for persistent storage. Implemented event streaming with Kafka to propagate pricing updates across regions with sub-second latency. Optimized the read path to sustain >3,000 reads per second, blasting out to hundreds of thousands of document reads/writes per second across the distributed cluster. Designed the system to gracefully degrade under extreme load with a worst-case latency well within 500ms and a p95 of approximately 30ms.
The Impact
Delivered a 6,000% throughput improvement, scaling from ~50 RPS to >3,000 reads/sec and >2,500 writes/sec. Reduced query latency from seconds to a p95 of ~30ms with worst-case well under 500ms. The system handled peak holiday loads without any degradation, protecting millions in revenue. The digital transformation modernized Hertz's core technology, enabling faster feature development and creating a platform capable of scaling to meet future demand across 10,000+ global locations.