Apr 2016 – Mar 2019 · 2 yrs 11 mos · Bengaluru Area, India

• As a key member of Amazon's Affiliate Marketing team, I contributed to critical systems for processing affiliate earnings at scale. The team built a scalable, fault-tolerant distributed event processing platform that calculated affiliate earnings by processing millions of daily events from Amazon's retail platform, correlating clicks, orders, and shipments driven through affiliate marketing channels. Within this ecosystem, I designed and implemented a real-time event ingestion system that subscribed to high-throughput Amazon digital retail events (processing millions of events per day), tracked digital returns through complex event correlation, implemented sophisticated filtering mechanisms for affiliate-specific returns, and built aggregation pipelines for affiliate-level return notifications integration into the earnings calculation workflow.
• The system leveraged AWS EMR (Elastic MapReduce) for large-scale data processing, running complex MapReduce jobs to analyze 60 days of historical data including click-through tracking, order completions, shipment confirmations, and return notifications, ensuring precise earnings calculations. Through deep analysis and optimization of EMR workflows, I demonstrated multi-million dollar annual cost savings in EMR infrastructure expenses while maintaining processing SLAs for the earnings calculation pipeline.
• During the team's migration from a monolithic payment processing system to a scalable microservice architecture, we encountered critical earnings calculation discrepancies between projected and actual affiliate payouts. Demonstrating customer obsession, I designed and implemented a large-scale reconciliation system that enabled automated detection and correction of earning mismatches. This solution significantly reduced customer escalations and resolve payment discrepancies, ensuring accurate affiliate compensation.

Aug 2012 – Apr 2016 · 3 yrs 8 mos · Bangalore Area, India

• As a Performance Engineer at Cognizant, I specialized in optimizing high-frequency trading systems and developing performance testing solutions.
• Working with critical trading systems for Indian Stock Exchange, I designed and implemented novel data structures and algorithms that reduced latency to 125 μs (90th percentile), while sophisticated journal data segregation methodology significantly improved trading engine throughput.
• I developed a Linux user-space BandwidthLimiter tool that enabled precise network traffic control at process and port level through TCP packet manipulation. This tool became instrumental in conducting resilience testing and performance benchmarking by allowing controlled simulation of various network conditions, bandwidth constraints, and latency scenarios.
• For a major banking client's trading application, I architected a load generation tool that simulated concurrent user behavior and trading scenarios. The tool integrated with the client's proprietary trading protocol, featuring dynamic request generation, configurable load patterns, and real-time performance monitoring capabilities.
• Improved the performance of trading cycle by adopting Google Protocol Buffers for efficient data serialization and adding strategic caching layers, we significantly improved the application's performance under high load conditions.
• In a notable innovation project, I prototyped a database virtualization tool that intercepted and analyzed MySQL network packets at the protocol level. The system featured TCP packet capture capabilities, query pattern analysis, customizable response injection, and acted as a standalone database server. The idea was to enable testing environments to operate without actual database servers and their costly licenses, as teams could simulate database interactions by modifying query-response patterns on the fly.