Bryan Tower

关于

Bryan Tower is a Principal Software Engineer at Microsoft working on the Augmented Learning and Reasoning team. His applied research work focuses on petabyte-scale data infrastructure, data science applications, graph analytics, and information visualization.  He applies machine learning algorithms to large data sets to discover trends and identify patterns of behavior.  He has applied experience in neuroscience, cyber-security, counter-human trafficking, fraud analytics, information retrieval, retail analytics, and real estate.

Bryan and his team are developing scalable vector indexing methods. These methods allow large language models to perform semantic search, which can find relevant results based on the meaning of the query. The team supports DiskANN (opens in new tab), a set of algorithms for approximate nearest neighbor search on large-scale vector data. DiskANN can handle real-time updates and simple filters while being fast, accurate and cost-effective.

Bryan’s team in the past has applied network machine learning to real problems. They have analyzed graphs with billions of edges. Through a pandemic initiative (Studies in Pandemic Preparedness – Microsoft Research (opens in new tab)), our network analytics can be applied to organizational science as detailed in this Harvard Business Review article (opens in new tab). This network machine learning research has also led to shipping recommendation engines within Bing, deployment of tools to combat tech fraud, and algorithms / tools to better understand news provenance.  Many of the supporting machine learning libraries have been open sourced in collaboration with Johns Hopkins University and are available as graspologic (opens in new tab) on GitHub.

Bryan’s team previously developed software for data science. They focus on the intersection of artificial intelligence and business intelligence for scalable interaction with data. The team helped enable text analysis within Power BI. The team shipped a suite of new data interfaces through Power BI. The open source visuals include the Network Navigator, Time Brush, Table Sorter, Attribute Slicer, Cluster Map, Facet Key, and Strippets Browser. The team has also shipped multiple “solutions templates” for social brand management and news analytics.

Bryan has helped fight phishing fraud by applying machine learning models to user behavior at large scale.  His work has been useful in tracking down and stopping fraudsters.

Prior to joining Microsoft, Bryan was a Senior Software Architect at Applied Technical Systems Inc. He led a variety of data science and big data efforts across several programs. He worked on revenue prediction models for new retail store locations. He also worked on revenue optimization through the use of big data for a brick and mortar retail chain.  He led several projects for the US Army.  Bryan focused on improving user workflows and improving the information retrieval experience.  He also worked on large scale graph analytic problems. He led a team that won a USDA innovation challenge competition.

Bryan’s graduate studies focused on Artificial Intelligence and Information Retrieval.