Olivia Buzek

Senior Staff Engineer - Trustworthy AI Tech Lead

IBM Research - watsonx.ai Platform

Nov 2021 - Present

• Unified IBM's fragmented Trustworthy AI research (fairness, explainability, robustness) into Watson Trust, a production-ready sklearn-compatible Python library. Previously these existed as separate codebases with incompatible data structures; the unified API made responsible AI practices accessible to product teams who had treated them as too obscure to implement.
• Founded IBM's Inner Source Program Office in 2021 and led the effort to coalesce 20 years of NLP development into the Python and Java-based Watson NLP library, which became the de facto standard for NLP across IBM in the pre-generative AI era. Won Outstanding Technical Achievement Award and IBM Corporate Award.
• Developed inner source practices which became a core component of IBM's technical strategy, included in the 2022 Global Technology Outlook, and a precursor to IBM's current open approach to generative AI.

Senior Engineering Manager

IBM Watson

May 2018 - Nov 2021

• Managed a team of 10 machine learning engineers across two cross-functional squads: one focused on ML algorithms, one on cloud productionization.
• Oversaw merger of two major NLP SaaS products, growing organization and onboarding remote teams into integrated matrix management.
• Stabilized post-acquisition team with high attrition and low morale over one year by reorienting organization to take advantage of new hires' extensive background in machine learning research.
• Created Algorithms Guild and Innovation Time to give ML insights a home on the product side, leading directly to a shipped beta feature: explainability for Natural Language Classifier.
• Developed inner source process for Watson NLP core language technology, allowing NLP and data science experts across IBM to volunteer contributions to the shared library.

Machine Learning Engineer

IBM Watson

Nov 2015 - Apr 2018

• Drove language expansion effort, ensuring NLP features reached non-English speaking users. Reduced time to add a new language by 75%. Won ISSIP Excellence in Service Innovation Award.
• Updated algorithms for NLP features including Keywords, Entities and Concepts; developed new Entities algorithm in TensorFlow.
• Built Python microservices for NLP features to replace legacy monolith.
• Gave internal talk "Linguistics for NLP System Builders" to help engineers understand linguistic concepts.

Forward Deployed Engineer

Palantir

Nov 2012 - Jul 2015

• Led technical outcomes for two large government client engagements.
• Maintained user-facing data analysis stack and built front-end data analytics solutions with Java, Oracle, and open source technologies.
• Improved Java-based data integrations with modern libraries, better error handling, and stability.

Graduate Student Researcher

Johns Hopkins University

Jun 2011 - Jun 2013

• Designed probabilistic graphical model based on LDA for text analysis and personality trait prediction.
• Built prediction system in Python for user schedules using Twitter data linguistic cues.

Undergraduate Research Assistant

University of Maryland Institute for Advanced Computer Studies

Nov 2009 - Jun 2011

• Developed algorithm in Ruby to identify difficult-to-translate spans in machine translation context.
• Improved runtime efficiency for exponential algorithm generating alternate sentence phrasings.

Johns Hopkins University

Computational Linguistics

Sep 2011 - Jun 2013

University of Maryland

B.S. Computer Science and Linguistics

Sep 2007 - Jun 2011

College Park Scholars Program