San Francisco Bay Area
I'm a software professional with 10+ years of experience, in both large organization and startups, building reliable and scalable systems, navigating collaboration challenges, and helping engineers grow.
Core Infrastructure Search Infrastructure
As part of a small Platform team in a dynamic and fast-paced startup environment, supported nearly 50 ML Research Engineers. Key Contributions: Pipeline Improvement: Led efforts to enhance pipeline efficiency for large scale training and validation dataset generation for Urban Pilot driving. Implemented optimizations to streamline the process and reduce turnaround time. Validation Framework: Spearheaded feature improvements for the validation framework, ensuring robustness and accuracy in evaluating model performance. Collaborated closely with Research Engineers to understand requirements and implement necessary enhancements. Diverse Tech Stack: Worked extensively with various technologies, including Milvus for similarity search, and Google Cloud Storage (GCS) and GCP Batch Workflows for data storage and processing.
Ads Org: - Improved reliability and efficiency of Ads Bidding Infrastructure - Directed product service for optimizing Ads audience spend trajectory - Rearchitected service for enhanced reliability and resource usage - Led multiple efficiency efforts resulting in 1 MW cumulative wins - Bootstrapped Infra support for ML-based bidding - Presented team’s vision to organizational leadership to acquire headcounts - Contributed to product launches, reliability, and dev efficiency efforts in Ad Conversion infra - Owned and drove goals with multiple cross-functional team alignments - Improved Push Reliability from 33% to 95% and reduced engineer’s time spent on push by 50% - Reduced average server downtime of a stateful sharded & replicated service during planned events by 90% AI Org: - Used data-driven approach to optimize host usage in Meta’s Inferencing platform - Prototyped implementation of using 0.35MW equivalent services in accumulated CPU hosts - Provided strategies to enhance reliability of services running under GPU hosts - Influenced improvements in quota management services across ~8MW of shared resources Others: - Reduced Local Android App size by 25%, ensuring compliance with Android app size limits - Supported Android camera API in various apps Overall: - Mentored 10+ Engineers and Interns - Conducted 100+ interviews in Coding and System Design
Assisted professors with labs and grading in courses: i) Object Oriented Design ii) Data Structures and Algorithms iii) Computational Thinking I iv) Software Engineering
Designed abstraction over HavoqGT (http://havoqgt.bitbucket.org/ ) for distributed random walk simulation. Experimented with temporal biased random walk simulation using various walk policies, e.g. random walk with restart, random walk with history, etc using curated Wikipedia Link Edit history as temporal dataset. Optimized history/states store and load for stateful random walk policies.