Dr. Tom Yeh
Tom Yeh is a professor of Computer Science and a recognized expert in Artificial Intelligence, with more than fifteen years of experience teaching, researching, and advancing applied AI technologies. He earned his PhD from the Massachusetts Institute of Technology (MIT), where he focused on vision-based user interfaces; a foundation that continues to shape his contributions to AI, computer vision, and human–computer interaction.
After completing his doctoral studies at MIT, Tom served as a postdoctoral fellow at the University of Maryland’s Institute for Advanced Computer Studies (UMIACS). In 2012, he joined the University of Colorado Boulder as an assistant professor in the Department of Computer Science, where he has since built a distinguished academic career and earned tenure. At CU Boulder, Tom teaches Artificial Intelligence, advises graduate researchers, and leads projects that bridge scientific discovery with real-world applications.
Tom’s research spans 3D printing, big data, computer vision, civic technology, and mobile security. He has authored more than thirty peer-reviewed publications across these fields and has received best paper awards and honorable mentions from leading conferences including CHI, UIST, and MobileHCI. His work has been funded by major federal agencies, including the National Science Foundation (NSF) and the Defense Advanced Research Projects Agency (DARPA), reflecting national recognition of the practical and scientific value of his research.
Throughout his career, Tom has focused on using AI and computational tools to solve complex societal and industry problems. His expertise and mentorship have supported the development of numerous student projects, early-stage technologies, and research collaborations.
In addition to his academic commitments, Tom plays an active role in commercial innovation, advising technology teams and contributing his expertise in AI system design and development. At home, he enjoys exploring the outdoors and spending time with his family.
Education
Ph.D., Computer Science, Massachusetts Institute of Technology (MIT)
B.Sc., Computer Science, Simon Fraser University
Designations
Professor of Computer Science, University of Colorado Boulder
Postdoctoral Research Fellow, University of Maryland Institute for Advanced Computer Studies (UMIACS)
What Tom built at QoEAgent
Most financial software assumes the data arrives structured. Below ten million dollars of revenue it usually does not. The bank statements are photocopies, and the general ledger is a spreadsheet somebody maintains by hand. Tom’s research area — vision-based interfaces and machine reasoning — is the part of computer science that addresses exactly that gap: getting a machine to read a document that was never designed to be machine-readable.
Tom is also known for teaching artificial intelligence by hand — working through neural networks and transformer mechanics step by step inside a spreadsheet, under the name AI by Hand. That turned out to be a more relevant background than it sounds. A quality of earnings engagement lives in spreadsheets: a trial balance, a general ledger export, a bank statement transcribed into rows that have to tie. Someone who has spent years making machine learning legible cell by cell in Excel already thinks about financial data the way a diligence professional does.
That work is the engine behind QoEAgent’s proof of cash and AI quality of earnings analysis. On a recent trailing-twelve-month engagement it read 431 scanned bank statements across 36 bank accounts, reconciled 29,823 general ledger transactions, and reached a 99.4% average match rate — leaving a single $32 variance flagged for a person to resolve.
The distinction matters more than it sounds. Tools that read a data room or tie out a filing start from documents authored to be read; the four categories of AI diligence tool get confused with one another constantly. Rebuilding cash from raw scanned statements is a different problem, and it is the one that has to be solved for the SBA quality of earnings requirement, where the cash proof must reconcile bank data to both the income statement and the tax return.
A credentialed professional still reviews the exceptions, forms the conclusions and signs the report. The automation is how the work gets done, not who is accountable for it.