Eben Holderness Email & Phone Number

Boston, MA, US

Waltham, MA

• Researched and designed methods for probing what cognitive/linguistic competencies large language models (LLMs) have access to via novel evaluation metrics and dataset development.
• Particular focus on ‘real world’ knowledge, which we probed using temporally linear text (and text+image pairs) structured as I/O event model graphs and evaluated in a question-answering task context.
• Contributed to the development of the CLAMS project (clams.ai), a Python-based open-source NLP toolkit, in collaboration with the Public Broadcasting Service (PBS) and the Library of Congress, with the aim of.

Psychosis Neurobiology Laboratory (Harvard Medical School)

• Co-wrote successful NIH R01 and R21 grants to apply machine learning to electronic health records for readmission risk prediction and analysis.
• Oversaw the full ML lifecycle from annotation guideline development to annotation task management (Brat and LabelStudio annotation tools) to model development (PyTorch, TensorFlow Keras, and SKLearn) to deployment.
• Our clinically interpretable models approached clinical expert parity.

LING 195: Introduction to Research in Linguistics and Computational Linguistics (Fall 2020)CS 114a: Fundamentals of Natural Language Processing (Fall 2022)

• Instructor of record for COSI 140: Natural Language Annotation for Machine Learning.
• Redesigned syllabus to cover the state-of-the-art methods in annotation, data augmentation, and machine learning.
• Selected feedback from students:– “The lectures were really clear and very helpful, and it was clear that the content we’re learning is very relevant and directly applicable to real–life projects that we might work on.

• Worked with a team of interns researching information architecture and improving the Pryon Information Retrieval - Question Answering (IR-QA) system.
• Implemented a semi-automated pipeline for targeted web scraping, hierarchically structured storage of these data in AWS S3, fine-tuning of the transformer-based retrieval model in AWS EC2, and delivery to customers.
• Our findings improved accuracy and reduced manual tuning required for domain adaptation.

Worked on the Inclusion Advisor system (https://www.workhuman.com/capabilities/inclusion-advisor/) to address and mitigate workplace bias. Using semi-supervised methods in vector space semantics, expanded model training size by ~50%. Developed a transformer-based architecture that significantly outperformed the existing LSTM-based model.

Aided in wrongful conviction investigations as part of the Justice Brandeis Law Project, contributing data-driven legal and archival research that led to the overturning of two individuals serving life sentences without parole.

Greater Boston Area

Semantic and syntactic annotations using brat rapid annotation tool

Brattleboro, Vermont

Interned at regional non-profit newspaper The Commons, reporting primarily on local issues and events. Edited and aided in design of the 2013 Autumn Supplement using Adobe Pagemaker.

Doctor Of Philosophy - Phd, Computer Science

Activities and Societies: Coordinator of Ethics in CS+LING Reading GroupAdvisor: James Pustejovsky Attending with full funding from the.

Master Of Science - Ms, Computational Linguistics, Gpa: 3.9

Activities and Societies: Coordinator of Ethics in NLP Reading GroupThesis Title: Improving Optical Character Recognition with.

Bachelor'S Degree (Cum Laude), Linguistics, Computer Science, Gpa: 3.7

Activities and Societies: Brandeis Rooftop Community Farm co-founder & manager, Brandeis Association of Music and Concert Organizers.

Education record

Activities and Societies: The Exonian (school newspaper) reporter, French Organ Scholar, Jazz Ensemble pianist & vibraphonist, Crew Team.

High School

Activities and Societies: School Newspaper editor-in-chief, National Honor Society inductee, Linguistics Society president, Hiking &.