Ming Zhao Email and Phone Number

Cloud Infrastructure Services, Services for Bare Metal Cloud

Network Services and Fleet Services in Bare Metal Cloud

I'm working on Google Managed VMs, which is a differentiating feature of Google Cloud. It combines the best features of both Infrastructure as a Service (Iaas) and Platform as a Service (Paas). My work involves building the underlying infrastructure including docker runtime,control plane (including deployment), and two important features: health checking (including AutoRecovery) and AutoScaling, see:https://cloud.google.com/appengine/docs/flexible/

Demand ForecastingIf a new product is released, it does not have history of demand and thus is very hard to forecast. I tackled the regression problemof forecasting, i.e. given ASIN attributes, forecast demand in the first 20 weeksof its life. I used random forest to solve this problem. The challenges are howto make it scale large data set, how to do feature selection using random forest,and how to optimize cost function using random forest. I’m the winner of themachine learning hackend 2012 for this track.Kindle eBook Content QualityI first built a platform to do quality check for all in-coming eBooks, andlarge scale backfilling and quality monitoring. This system has processed millions of eBooks. Secondly, I improved the language classifier by usingwikipedia data. By mentoring my intern, we not only improved the languagecoverage, but also improved the recognition accuracy.Thirdly, I built a system to estimated the number of pages in Kindle eBooks. The system involves several Amazon teams. I designed and iterated severalversions of the algorithms to estimate the number of pages, and implementedthe system to integrate with other teams. The estimated number of pages hasbecome a critical feature for one the retail website of Kindle eBooks. I also mentored an intern to do photo and line-art classification with high accuracy.Fourthly, I designed and built a service to do typo detection.Kindle eBook Bulk IngestionI worked on converting physical books to Kindle eBooks. Most of my work isrelated to image processing and OCR components of the Kindle eBook bulkingestion pipeline, for example, image segmentation, image deskew, image despeckle, image enhancement, image page alignment, OCR with Abbyy, Nuance, and Tesseract engines.Silk Browser in AWSI also work in Amazon silk browser team in AWS to make browsing faster. It involved theAWS cloud computing and the Android device development.

YouTube Video ClassicationGoogle Landmark Recognitionhttp://googleblog.blogspot.com/2009/06/new-landmark-in-computer-vision.htmlGoogle Celebrity Recognition: http://www.informationweek.com/news/internet/google/showArticle.jhtml?articleID=217600713