Donald Bertucci

I am a senior at Oregon State University studying Computer Science and Mathematics.

I'm fascinated in how we can use visualizations to peer into the complex patterns and emergent behavior of machine learning models.

I've worked on visualizing neural networks and large-scale data as a member of the CMU Data Interaction Group and previously as a member of the Data Interaction and Visualization Lab at Oregon State University.

person donnybertucci.com mail bertuccd@oregonstate.edu code github(@xnought)

Education

2020 – Present

B.S. Computer Science

Oregon State University
Degree focus on Artificial Intelligence
Minor in Mathematics

Experience

Summer 2023

Carnegie Mellon University

Research Intern, CMU Data Interaction Group (DIG)
Researched interactive methods to improve language model prompt generation and transparency with Adam Perer . Developed interactive visualizations of neural network compression/quantization error [_].
Sep. 2022 – Jun. 2023

Carnegie Mellon University

Research Assistant, CMU Data Interaction Group (DIG)
Developing human-centered ways to evaluate Machine Learning model behavior within Zeno [_]. Enabling linked visualizations at scale with FalconVis [_] with Dominik Moritz .
Summer 2022

Carnegie Mellon University

REU Intern, HCII Summer Undergraduate Research Program
Developed user interfaces to interactively discover poor behavior in neural networks [_].
Advised by Ángel Alexander Cabrera and Adam Perer . Hosted by the Data Interaction Group (DIG) .
Aug. 2021 – Jun. 2022

Oregon State University

Research Assistant, Data Interaction and Visualization (DIV) Lab
Developed user interfaces to visualize large data and interpret complex machine learning models [_, _]. Mentored and advised by Minsuk Kahng .
Summer 2021

Oregon State University

EECS REU Summer Cohort
Program dedicated to improving research communication and presentation skills. Led by Alan Fern .
Feb. 2021 – June 2021

Oregon State University

REU Intern, URSA Engage Research Program
Developed interactive interfaces to visualize difficult concepts in learned neural networks [_, _]. Advised by Minsuk Kahng .

Publications

Conference

C3

Zeno: An Interactive Framework for Behavioral Evaluation of Machine Learning

Ángel Alexander Cabrera , Erica Fu , Donald Bertucci , Kenneth Holstein , Ameet Talwalkar , Jason I. Hong , and Adam Perer
ACM Conference on Human Factors in Computing Systems (CHI). Hamburg, Germany, 2023.
link Link code Code description PDF
C2

DendroMap: Visual Exploration of Large-Scale Image Datasets for Machine Learning with Treemaps

Donald Bertucci , Md Montaser Hamid , Yashwanthi Anand , Anita Ruangrotsakun , Delyar Tabatabai ,
Melissa Perez , and Minsuk Kahng
IEEE Transactions on Visualization and Computer Graphics (IEEE VIS 2022). Oklahoma City, OK
link Link play_arrow Video code Code description PDF
C1

Beyond Value: CHECKLIST for Testing Inferences in Planning-Based Reinforcement Learning

Kin-Ho Lam , Delyar Tabatabai , Jed Irvine , Donald Bertucci , Anita Ruangrotsakun , Minsuk Kahng , and Alan Fern
32nd International Conference on Automated Planning and Scheduling (ICAPS 2022).
description PDF

Workshop

W2

Backprop Explainer: Interactive Explanation of Backpropagation in Neural Network Training

Donald Bertucci and Minsuk Kahng
Workshop on Visualization for AI Explainability (VISxAI, IEEE VIS 2021).
link Link play_arrow Video code Code
W1

An Interactive Introduction to Autoencoders

Donald Bertucci
Workshop on Visualization for AI Explainability (VISxAI, IEEE VIS 2021).
link Link play_arrow Video code Code

Research Projects

P4

Visualizing Neural Network Compression

Donald Bertucci and Adam Perer
An interactive article exploring how model compression error affects neural network behavior.
link Link code Code
P3

FalconVis: A Library to Cross-Filter Billions of Data Entries on the Web

Donald Bertucci and Dominik Moritz
A JavaScript library for visualizing big data on the web with your custom visualizations and scalable data formats.
link Link code Code
P2

Finding the Distance Function in the Poincaré Disk using Stereographic Projection

Donald Bertucci
A paper that derives the Poicaré disk distance function using stereographic projection from Minkowski Space.
Non-Euclidean Geometry with Tevian Dray, Oregon State University (2023). Corvallis, OR
description PDF
P1

Mirror: Interactive Discovery of Blindspots in Machine Learning Models

Donald Bertucci , Ángel Alexander Cabrera , Nari Johnson , Gregory Plumb , Erica Fu , and Adam Perer
An interactive tool to discover errors in machine learning models. Detailed in the poster and implemented in Zeno.
Human-Computer Interaction Institute (HCII) Summer Research Showcase (2022). Pittsburgh, PA
description PDF

Presentations

  • Mirror [P1 ] at CMU HCII Summer Research Program Showcase
  • Dendromap [C2 ] at the Visualization Conference (IEEE VIS 2022 in Oklahoma City, OK)
  • Interactive Autoencoder [W1 ] at workshop on Visualization for AI Explainability (VISxAI, IEEE VIS 2021)
  • Backprop Explainer [W2 ] at workshop on Visualization for AI Explainability (VISxAI, IEEE VIS 2021)
  • Backprop Explainer [W2 ] at Celebrating Undergraduate Excellence (CUE 2021 at Oregon State University)
  • Backprop Explainer [W2 ] at OSU 2021 EECS Undergraduate Summer Cohort

Service

References

Dr. Minsuk Kahng

Research Scientist at Google People + AI Research (PAIR)
Previously Assistant Professor at Oregon State University
minsuk.com

Dr. Adam Perer

Assistant Professor at Carnegie Mellon University
Co-Lead Data Interaction Group at CMU HCII
perer.org

Dr. Dominik Moritz

Senior Project Scientist at Carnegie Mellon University
Co-Lead Data Interaction Group at CMU HCII
Research Scientist at Apple Machine Intelligence visualization team
domoritz.de

Alex Cabrera

PhD Student at Carnegie Mellon University
ZenoML
cabreraalex.com