Jan 2024 – Aug 2024 · 7 mos

• Architected and developed key aspects of Heimdall, a high-performance, extensible rule-based throttling/filtering framework designed for seamless integration with other services. Heimdall leverages dynamic predicates and caching to filter and block requests based on customer attributes, achieving a P99 blocking latency under 3ms. I spearheaded the design and coding to ensure extensibility and low latency, utilizing passthrough filters for efficient processing and integrating caching mechanisms for improved performance. Furthermore, I established CI/CD pipelines for efficient development and deployment, configured dev, staging, and sanity environments for rigorous testing.

Jul 2021 – Dec 2023 · 2 yrs 5 mos

• Optimized Block Entity System: Reduced leakages, improved performance, and minimized downtime through proactive maintenance and optimization. This system oversees real-time traffic blocking based on 24 million rules with sub-millisecond execution time, handling a massive 150+ billion daily transactions.
• Enhanced Creative Review System (CRS): Streamlined creative review workflows by working with third-party systems, enabling processing of 5 million unique creatives monthly. Additionally, implemented filters for the massive 150+ billion daily traffic flow.
• Scalable Rule-Based System: Optimized a client-server system with over 10 million rules, achieving 1ms latency for blocking requests at a 100 million TPS rate.
• Optimization with Lost Opportunity System: Designed and implemented a high-performance caching solution (Lost Opportunity System) to identify optimal blocking entities for response blocking strategies. This application-level cache utilizes a sliding window and frequency-based approach to deliver lists with millisecond latency for 100 million TPS, ultimately driving profit optimization.
• Improved Database Performance: Successfully identified and optimized critical Block Entity System queries that were causing server freezes. Recalibrated indexes and queries, resulting in a significant reduction in query time from minutes to seconds.

Feb 2021 – Jul 2021 · 5 mos

• Worked with the AdExchange Team, primarily on a product to reduce the outbound traffic for saving infra cost.
• Worked on redesigning the exchange platform's logging setup to reduce the log's size by 40% and CPU usage. and build a log collector to persist logs in case of failures in Google Cloud storage.
• Worked on the timeline feature to record the timeline of a request flowing thought the exchange and its micro services.