# Joseph W. > Software Engineer for ML Infra/Platform Location: San Francisco, California, United States Profile: https://flows.cv/josephw Architect for several large-scale enterprise systems powered by big data analytics and machine learning algorithms. ## Work Experience ### Principal Engineer @ Uber Jan 2015 – Present | San Francisco Bay Area TL for Michelangelo - Uber Machine Learning platform. ### Principal Architect (director equivalent position) @ Walmart eCommerce Jan 2013 – Jan 2015 | San Bruno • Hand-on chief architect for an automated price matching system SavingsCatcher that leverages no-sql technologies such as HBase and Cassandra. The project improves store sales and increases Walmart.com traffic & conversion accounting for 9-figure revenue in 2014 and 2015. • Created machine learning models and processes for detecting potential fraudulent activities that reduced cost by millions of dollars. • Developed leader-board system and low-price item suggestion system to increase customer engagement. • Implemented product grouping system by text mining descriptions and product categories. • Evaluated new Hadoop technologies. Served on internal Hadoop advisory board. • Filed five patents. ### Search Science Architect (director equivalent position) @ Shopzilla Jan 2011 – Jan 2013 | Greater Los Angeles Area • Developed machine learning models including a query-result relevance model with high predictive power on click-through rates (using algorithms like conditional inference tree, boosting, and logistic regression), an advertiser conversion model and an ad click-through predictive model for SEM traffic (using decision trees), a real-time page quality scoring system, a low quality merchants detection algorithm, an ensemble method for advertisement classification (using ensemble method that uses SVM, NN, etc. & tools like VW), a Solr query rewrite system for improving relevance, price conversion predictive models, etc. Wrote a query recommendation system and a category suggestion system (using collaborative filtering approach). • Developed a keyword bidding system that improved monetization and conversion. • Chief architect for a scalable event-queue driven keyword scoring platform for multiple countries on a 128-node Hadoop cluster. It improved throughput over previous system by at least 10 times. • Designed and implemented data mining pipeline and search quality monitoring system. ### Head of Research and Development @ LeadPoint Jan 2010 – Jan 2011 • Served as the chief architect for a new consumer centric offer platform. Was able to deliver the beta release four months after the initial design. • • Developed new algorithms for managing offer quality and advertisers spending. • Written a new serving system that improved the average response time to less than 30ms, at least 130 times better than the existing products. ### Director/Architect @ Adknowledge Jan 2009 – Jan 2010 Chief architect for the Domains group. • Led a team to implement new domain product that went production ahead of schedule. The annual revenue already exceeded eight million dollars. • Developed patent pending technologies for navigation and advertiser bid discounting that improved monetization for publishers and decreased cost for advertisers. • Developed a domain traffic quality prediction algorithm using various machine learning techniques. • Designed a statistical feedback model for pricing publishers’ traffic that achieved the best overall channel quality. • Implemented a search-query to semantic concept mapping system that increased query coverage. • Designed and implemented a behavior targeting platform for sharing users information across multiple publishers and ad serving products. • Developed an algorithm for detecting click spam traffic. • Worked with business unit on project priority and on developing unique competitive features. • Coordinated with other groups on resource allocation and capacity planning. ### Researcher, Search & Advertisting Science @ Yahoo! Jan 2008 – Jan 2009 | Burbank, CA - Worked with Yahoo! Research to develop a query rewrite system that uses featues from web pages. Bucket test showed 2.1% RPS lift. - Worked on unified query rewrite generation pipeline on the grid. - Analyzed and proposed solution for query coverage issue for an ad listing serving subsystem. ### Principal Engineer, Domain Match @ Yahoo! Jan 2006 – Jan 2008 | Burbank, CA Architect for the Domain Match group for Yahoo! Search Marketing. - Evaluated and created tokenization service for French, German, and Spanish. - Designed and implement a term optimization system Nitro that improved STR by at least 20% and RPS by at least 33% for the US market. [See US patent 7647316 B2.] - Developed a blend-model for terms suggestion that improved relevance score by 39% & RPS by 12% for the US market, and RPS by 11% for the France, Germany, and Spain. - Worked with Yahoo! Research to develop phrase categorizer for taxonomy. - Researched domain concept detection using searchbox queries to improve overall relevance of terms suggestion. Initial bucket test result showed improvement of STR by more than 20% for low STR domains. - Acted as the research liaison between development and Yahoo! Research & Matching Science. Led and coordinated all research activities within the Domain Match group. - Served as a technical lead and architecture on several key projects for the Domain Match group. ### Architect/Senior Software Consultant (part-time) @ Focus Simulation Jan 2006 – Jan 2007 ### Senior Consultant @ Reuters Jan 1998 – Jan 2005 Key architectural member of the Server Group. - Researched and prototyped next generation server using EJB, Java servlets, JSP, Oracle application server, Tomcat, SOAP, and WRAP, Apache, and IIS. - Designed and implemented high performance web server and desktop application that supported failure recovery, files sharing, and entitlement policy. - Designed and implemented proxy for web service gateway. - Designed and implemented web services that cached market quote information in order to provide fast query response. - Analyzed server & network issues and provided fast solutions that improved system reliability - Served as a key technical contact for numerous financial institutional clients. - Coordinated rapid development efforts for several high performance real-time systems. Designed and delivered quality software products to clients ahead of time, which allowed clients to achieve their roll out schedules. ### Principal Engineer @ Teradata Jan 1993 – Jan 1998 | El Seguno, CA Member of the Parallel Database Extension (PDE) kernel group responsible for the design and development of kernel level drivers for supporting large parallel database application on UNIX MP-RAS and NT. - Developed automated stress test software that was used to identify system bottlenecks and ensure system scalability and stability. - Improved virtual processor performance and eliminated lazy panic codes. - Implemented file segment subsystem for database application. - Recipient of various company divisional awards for achieving high customer satisfaction, teamwork, and reaching production goal ahead of schedule. ### Senior Consultant (part-time) @ Candle Corporation Jan 1996 – Jan 1997 ### Teaching Assistant (part-time) @ University of California Jan 1994 – Jan 1997 ## Education ### PhD in Information and Computer Science UC Irvine ### MS in Financial Engineering and Management Drucker School of Management ### Stanford University ## Contact & Social - LinkedIn: https://linkedin.com/in/josephwangphd --- Source: https://flows.cv/josephw JSON Resume: https://flows.cv/josephw/resume.json Last updated: 2026-04-12