Portrait of Yi-Min Wang

Yi-Min Wang

Corporate Vice President, Business AI and Research


In May 2018, I was named Chief Technology Officer, Artificial Intelligence (CTO, AI) of Microsoft’s Business Applications Group.

I became Corporate Vice President, Business AI and Research, in March 2017. My organization focuses on AI-powered solutions for enterprise digital transformation, including Customer Care Automation, Sales and Marketing Optimization, Demand Forecasting, Financial Predictions, and Machine Reading Comprehension (MRC).

In early 2017, I led the acquisition of deep-learning start-up Maluuba.

I was named a Distinguished Scientist of Microsoft in March 2015 for being “recognized internationally for his contributions to dependable computing and web security” and for “establishing a company-wide reputation for both effective and creative management”. In November 2015, I gave a keynote speech at the UIUC ECE Launch event to summarize the four things I learned from my 22-year career (see video, fast forward to the 11:40 mark).

I became the Managing Director of the newly created MSR Technologies (MSR-T) organization in January 2014. At MSR-T, we focus on turning world-class research into world-changing innovation. Our investment areas include Input Technologies (e.g., world-record texting), Deep Learning (both algorithms and scale), Natural Language Processing (e.g., Skype Translator and Cortana), Wearables (e.g., HoloLens and Microsoft Band), and Distributed Systems.

I became a Deputy Managing Director for Microsoft Research-Redmond (MSR-R) in August 2011. Together with Peter Lee and Eric Horvitz, we formed the Office of Directors (OOD) which supports all 300+ researchers and engineers in the Lab towards the goals of Agility, Diversity, and Strategy. In November 2012, Eric and I became the Managing Co-Director of the lab when Peter expanded his responsibility to manage all MSR US Labs.

Between 2012 and 2014, I managed the MSR Gaia Skunkworks Program, which produced the following three world-record successes: (1) the FDS team broke two MinuteSort records – see article and record page; (2) the Blackbird team broke the Guinness World Record on fastest texting and set a new record on blind texting – see article; (3) the Adam team set a record on ImageNet 22K object recognition top-1 accuracy – see article and OSDI paper.

I was previously Director of ISRC (Internet Services Research Center), an R&D organization dedicated to developing technologies for Search, Ads, and Online Services. Between 2007 and 2011, ISRC focused on web-scale data-related technologies including Search Quality Diagnosis and Metrics, Web-Scale Language Model, Highly Interactive Dialog Model, Structured Data (e.g., Wolfram|Alpha answers), and Scalable Dynamic Crawling. Personally, I played the role of Director of Search Quality for the Bing team. I invented the Automated Relevance Diagnosis System (ARDS) which took hundreds of thousands of user dissatisfaction reports and performed automated and systematic diagnosis to identify the Search component that is responsible for causing each of the dissatisfactions. ARDS has been a critical part of Bing team’s search quality process and is widely recognized for its contribution to Bing’s significant quality improvement from 2008 to 2011.

Prior to the creation of ISRC, my research work was primarily in the areas of dependable computing and web security. I was elected to IEEE Fellow for my contributions in those two areas. I had published extensively on checkpointing and rollback-recovery and was a main co-author of the most influential survey paper on that topic. In 2005, I invented Strider HoneyMonkey – the first automated system to patrol the Web and hunt for malicious websites that exploit zero-day vulnerabilities. The HoneyMonkey technique has become the de facto standard for both the security industry and the search engine industry.

In 2007, I invented Strider Search Ranger – the first search-spam detection system based on dynamic crawling and traffic analysis. The work was featured in the New York Times and has had an industry-wide impact on wiping out search-spam.