Haritha Sekar

Architected and built a scalable microservice platform (Mars 2.0) using Spring Boot, Docker, and AWS ECS/Fargate, decomposing a monolithic system into autonomous services — reduced operational complexity and improved deployment velocity. Developed an autonomous infrastructure provisioning system using AWS CDK that automated network setup, DNS configuration, and resource creation — cut service setup time by over 50%. Designed an API-first development workflow using OpenAPI specifications and Swagger Codegen, enabling teams to generate service interfaces from YAML definitions and accelerating onboarding for new widget developers. Led content ranking optimization experiments that tested updated ML model features (customer propensity, widget preference signals) and improved training pipelines (outlier elimination, rebalanced training data) — drove $150M+ in incremental revenue.

Implemented personalized ranking features using customer intent signals, designing a novel widget-relevance feature generator that incorporated complete shopping journey data instead of single-mission signals. Ran and analyzed multiple A/B tests across global marketplaces (US, EU, JP, IN), making data-driven launch decisions including strategic dial-downs around peak traffic events (Prime Day). Built comprehensive monitoring and logging using Micrometer, SLF4J, and Cloud Watch, enabling real-time diagnostics across distributed services. Automated weekly business review (WBR) report generation, eliminating manual workflows and increasing attribution accuracy by 1–3% — won 'Highest Impact Potential' at org-wide hackathon.

Reduced data pipeline processing time by 50%, saving $26K/year in infrastructure costs by identifying and relocating misapplied log filters. Took ownership of operational excellence: established on-call processes, drafted team SLAs, and represented the team in quarterly org-wide OE reviews. Mentored 3+ junior engineers and new hires, creating onboarding documentation and reducing ramp-up time. Conducted technical interviews for SDE hiring.

Built end-to-end recommendation widgets for Amazon homepage (desktop & mobile) that generated $34.8M in cumulative revenue across 8 A/B experiments, with 6 successfully launched to production. Designed a location-based recommendation engine using collaborative filtering on geographic shopping data, integrating real-time customer location services — generated $2.7M in revenue and expanded to international markets.

Created a private-brand recommendation system using exponential decay modeling on customer browsing behavior, surfacing relevant substitutes — drove $7M in revenue. Engineered a reusable multi-collection recommendation framework serving personalized product collections at scale with p99 latency under 140ms, enabling other teams to build diverse recommendation experiences.

Improved recommendation quality by developing a cosine-similarity filter on item embeddings, reducing irrelevant results by 30% while maintaining high coverage — resulting in $18M+ in attributed revenue. Mentored a summer intern end-to-end: authored the project proposal (selected org-wide), created technical documentation, and guided project to production deployment.

Built a brand discovery system using Word2Vec NLP and cosine similarity to identify top-searched brands and items per product category — completed ahead of schedule, surpassing all milestone targets.