Sainath Vellal

Principal Research Engineer
Sunnyvale, California

Sainath has been a research engineer with Yahoo! since the past 8 years and has been building systems and algorithms that touch and have had an impact on almost every major media property across Yahoo!. Some examples of the systems that he led the prototyping/implementation/deployment for Yahoo! Media include

  • Content Trends - a system that generates trending entities as a signal and dynamically weaves/generates article streams every 10 mins. This system scales to multiple segments (politics, sports, finance…) and across languages. Currently deployed on the Yahoo! Search App - Trending Now news stream. Also available on Yahoo! Mobile Search - Sports. Will soon be available on Yahoo! Homepage - www.yahoo.com
  • Story Clustering - a system that organically/algorithmically generates related stories for news articles in real time. This was deployed on the Yahoo! homepage.   

During his tenure with Yahoo! Mail, Sainath led a project to detect automailers and tag them.

Sainath has also worked on problems in Yahoo! Search during the bootstrapping days to create seeds and feeds for content acquisition for the Yahoo! Crawler framework. He was also the previous maintainer of (YCT - Yahoo! Content Taxonomy) and worked on developing a pipeline for Active Learning to improve the YCT classifier category performance (Human in the middle approach). 

In his previous life as a developer, Sainath worked on Machine Learning inference libraries which are used to autolabel content that flows through the huge content ingestion systems at Yahoo!. He has built and deployed to production many modules for CAP (Content Analysis Platform) including NER (Named Entity Recognition) and other NLP modules. 

Building internet scale AI systems and platforms using latest ML/NLP infrastructures and Data mining algorithms is his primary focus.


Content Understanding, Data Science, Information Retrieval, Machine Learning, Natural Language & Dialogue Understanding, Scalable Systems