Avik Paul
Software Engineer @Google
Signup · Get unlimited contacts
WORK HISTORY
Software Engineer @Google
Google Inc.(Mountain View, CA) 2010 - present ]- ML Infrastructure Engineer in Google Analytics team. High throughput low latency pipelines in C++. Resource efficient designs. Exceptional fluency in C++14, C++17. Dabble in C++20 (limited features). C++ Readability mentor at Google. Infrastructure: storage, microservices, various paradigms of distributed pipelines (task based, flume/mapreduce). Expertise in improving throughput of C++ jobs up to 20X from legacy implementations at no extra resource (w/ extensive multithreading), and cutting down resource cost. Build petabytes sized storage and ETL pipelines that prepare data for ML training and serving. ML Engineering : Experience in engineering and infra aspects of ML - training data generation, feature engineering, skew reduction, sampling, reference selection, forward validation techniques, monitoring. Built feature-engineering frameworks and distributed serving stack from scratch which powers several predictive features in Google Analytics : Predictive audience Smart-Goal, Session-Quality, Probability of Conversion, etc.[Google Bangalore (India): 20••••09 ]- Areas include extensive distributed programming (mapreduce) and data mining, spam classification in social network, prevention and filtering pipelines, large scale processing of massive social graph data. Automated classification and clustering of users and content based on various signals. High volume spam removal pipelines, pipeline to detect trending content, URLs, and XSS attacks. Machine learning, AI. Data mining. Backend development - C++, STL, Java, Python and other scripting languages. Short experience in UI development with cutting edge Google technology (GWT).
EDUCATION
Jadavpur University
B.E.
Indian Institute of Technology, Kanpur
M.Tech
SKILLS
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.