Leading the Search Core Engineering team. We are responsible for the platform that powered the search, personalization, recommendations, and related services on YP.com, m.yp.com, YPMobile apps, and YP AdNetwork. We are also responsible for relevance and performance improvements in our Search Retrieval and Ranking Algorithms.

Cloud Computing (2013.9 —— 2016.4)OpenStack is an open source software for creating and managing clouds. The task of our team was to make OpenStack the core infrastructure for Yahoo. My focus is on Baremetal provisioning. As a tech lead, I patched OpenStack code to integrate with Y! core systems and to support the Y!-scale; I optimized nova commands to allow faster provision, and to reduce error rates. I was able to reduce the nova quota-show command response time from 2 mins to 2s with optimization, and decreased failure rate from 35% to 3.57%. I worked with Ops team to resolve user issues by collecting information from different data sources, and investigating the root causes. Apart from supporting 75% new hardware provisions with OpenStack Baremetal, I also led the efforts for migrating all Yahoo physical hosts onto OpenStack transparently, without impacting live traffic.

Personalization System (2012.4 —— 2013.8)The CORE+ platform was the improved version of CORE platform. It provides personalized content recommendations on Yahoo homepage with more than 2 Billion daily impressions. As one of the key contributors in personalization team, I worked on components including scoring engine, content ingestion, data processing pipeline, and backend model training framework. After Yahoo’s new homepage launch, the stream is up 54% in weekly clicks per unique user. Due to our excellent work, both personalization team and homepage team got recognized with Yahoo SuperStar Awards in 2013.

Revenue Optimization through Traffic Shaping (2010.10 —— 2012.3)I was the lead developer and owner of Traffic Shaping project. Traffic shaping optimized both click through rate and the revenue generated on the article landing page after user clicks it from Yahoo homepage. It involved content recommendation platform, 3 different ads serving systems, and ads revenue impact evaluation system. I collaborated with Ads systems team and scientists in Yahoo lab since the project utilized data from different sources including content serving logs, ads serving logs, and revenue data for each ads. The bucket test on Yahoo homepage displayed 3-6% revenue increase!Recommendation Systems (2005.10 —— 2010.9)I was the only engineer to work with researchers to power MetroPA module nd Taiwan Mall property of CORE. CORE powered the large-scale recommendation modules on various Yahoo properties. Product features included popularity based recommendation, light recommendation (demo segment based), deep personalization (per user inferred interests). Yahoo Frontpage reached a 270% increase in engagement due to the CORE optimizationI was responsible for the modules deployed on Realestate, Shopping, Travel, BUZZ and IMS of VIBES. VIBES was an item recommendation platform that support affinity based collaborative filtering and attribute-based similarity calculation. It extracted user activity data, applied filters, converted data into the right format, then fed data into model training script. VIBES recommendation modules were deployed on various popular Yahoo properties.

Designed and implemented language constructs and compiler optimization techniques to support high performance data-intensive applications on the Grid or distributed environment, System Research Group, CSE Department, Ohio State University, OH.

+ Lecturer for CSE459.22 Programming in C+++ Grader for CIS755 Programming Language