Agnish Upadhyay

– Implemented a resilient job scheduler library using Asynq, Kafka & PostgreSQL architecture that transformed monolithic batch processing into granular, independently retryable tasks, enabling 50% resource reduction and eliminating cascading failures across workloads. – Designed and implemented a scalable event-driven data pipeline processing 600M+ ad interaction events daily via Kafka, with ClickHouse as the backbone to deliver near real-time analytics to power a funnel view of business metrics such as spends, revenue, CTR, RoAS. – Worked on integrating User Targeted Promotions with Ads (industry-first initiative) by coupling user segment based pricing with ads campaign management, resulting in a 13% increased ad CTR and 25% increase in RoAS. – Led the development & integration of Real Time Cross Selling ad experiences in Zepto’s consumer app on Search, Cart & Product Details Page, improving CTR by 11% and ad revenue by 8%. – Enhanced monitoring and observability by instrumenting key service metrics with Prometheus and CloudWatch, and visualizing them in Grafana dashboards, enabling faster incident response and improved error traceability across services.

– Implemented Fuzzy Matching using NLP after cleaning Economic Times data to match 4.5M user entries with a set of acceptable choices in order to improve the data quality by standardising the user entries. – Utilized a combination of vectorization, cosine similarity, and Levenshtein’s distance for accurate matching.