Ivan Brugere
Lead AI Research Scientist  ·  JPMorgan Chase & Co.  ·  Chicago, IL
25+
Publications
12
Patents
6+
Years Industry
4
Top Venues

I am a Lead AI Research Scientist at JP Morgan with 6 years of experience focusing on perturbation robustness, adversarial robustness, and privacy-preserving ML. I have 25+ publications, many in top venues, and 12 filed patents. I leverage an interdisciplinary background, domain adaptability, and a balanced track record of publications, patents, and deployed systems for high-impact research.

My work spans diverse domains including biology, ecology, finance, and legal — I deeply value interdisciplinary partnerships and articulating high-impact computational problems in new areas. During my current role at JP Morgan, I focused on explainable and auditable models in highly regulated applications, and robust methods with respect to model and data perturbation. My PhD (defended 2020) focused on graph topology inference; graph learning remains a common lens for how I formulate problems.

I am seeking industrial AI scientist roles where I can formulate novel, robust AI methods across large and classical ML models in unique and challenging domains, with an opportunity to mentor and develop emerging researchers.

Get in touch
Ivan Brugere

Experience

J.P. Morgan Chase & Co.
Jan. 2021–Current
Lead AI Research Scientist Promoted Jan. 2025 — Trustworthy AI
Working on AI robustness and fairness: explainable and robust LLM methods, fair agent-based learning, robust and fair tree-based ensembles. Published in ICML, NeurIPS, TMLR, EMNLP, 12 patents filed.
  • Co-led research initiative on perturbation robustness for tree-based models, leading to 2 NeurIPS publications and 2 patent filings.
  • Designed and deployed a model perturbation and auditing system for fairness, targeting firm-wide consumer-facing models (patent filed).
  • Developed novel Bayesian model search frameworks for fair mortgage and auto loan pricing (patent filed).
  • Led research of two summer AI research interns, yielding 1 NeurIPS, 1 TMLR publication, and 1 successful full-time scientist conversion.
  • Co-led external research collaborations with university faculty at NYU, MSU, UMD, Harvard.
Salesforce
Jan. 2019–March 2020
Research Scientist — AI For Good
  • Collaborated with Salesforce non-profit customers and external collaborators on fairness-aware AutoML applications.
  • Published research on graph augmentation for equitable access (AAAI AIES 2021), 1 patent filed.
Amazon
June–October 2018
Applied Scientist Intern
Mentor: Alex Smola
Designed deep learning graph APIs on MXNet for scalable training as part of the DGL project. Part of the Amazon Web Services AI Platforms team.
Microsoft
May–August 2015
Data Science Intern
Mentor: Marcello Hasegawa
Developed textual and device models for novelty detection and attribution in the Windows 10 user population.
Lawrence Livermore National Laboratory
May–August 2014
Research Intern
Mentor: Brian Gallagher
Formulated graph inference problems over several scientific research domains.
University of Illinois at Chicago
2013–2015
Research Fellow
Mentor: Prof. Venkat Venkatakrishnan
Graph-based models for attribute inference and privacy preservation on real mobile device datasets.
Technicolor Research
May–August 2013
Research Intern
Mentor: Fernando Silveira
Rule discovery for biometric sensor time series data for actionable analysis of film audiences. Developed methods to discover and visualize dynamic audience communities.
University of Illinois at Chicago
2012–2017
Research Assistant
Advisor: Prof. Tanya Berger-Wolf
Model selection for graph structure inference and prediction. Focusing on ecology and population biology domains.
University of Minnesota
2010–2012
Research Assistant
Advisor: Prof. Vipin Kumar
Time series change detection and anomaly detection on large remote sensing datasets. Focused on incorporating spatial aspects for change significance testing, and domain-driven information retrieval.
University of Minnesota
2004–2007
Web Applications Developer
Department of Computer Science.
University of Minnesota
2002–2003
Web Applications Developer
College of Liberal Arts.

Education

2012–2020
Ph.D., Computer Science
University of Illinois at Chicago — Advisor: Prof. Tanya Berger-Wolf
Thesis: Network Structure Inference: Methodology and Applications
2009–2012
M.S., Computer Science
University of Minnesota — Advisor: Prof. Vipin Kumar
Thesis: Approximate Search on Massive Spatiotemporal Datasets
2007–2009
M.A., International Affairs
The New School
2002–2007
B.S. Computer Science  ·  B.A. Cultural Studies & Comparative Literature
University of Minnesota

Publications

Full publications and patents at: Google Scholar
Selected Publications
Calibrating LLM Confidence by Probing Perturbed Representation Stability
Novel method applying adversarial perturbations to LLM hidden states to assess confidence via internal stability, reducing calibration error by ~55% across multiple models.
R. Khanmohammadi, E. Miahi, M. Mardikoraem, S. Kaur, I. Brugere, C. Smiley, K.S. Thind, M.M. Ghassemi
EMNLP 2025
Cross-Domain Graph Data Scaling: A Showcase with Diffusion Models
Introduces UniAug, a universal graph augmentor using discrete diffusion models pre-trained on thousands of graphs to enable cross-domain data scaling and adaptive structure enhancement for downstream tasks.
W. Tang, H. Mao, D. Dervovic, I. Brugere, S. Mishra, Y. Xie, J. Tang
NeurIPS 2025
Interpretable LLM-based Table Question Answering
Introduces Plan-of-SQLs (POS) method decomposing table queries into transparent SQL steps for interpretable reasoning, achieving competitive accuracy with 25× fewer LLM calls.
G. Nguyen, I. Brugere, S. Sharma, S. Kariyappa, A.T. Nguyen, F. Lecue
TMLR June 2025
RashomonGB: Analyzing the Rashomon Effect and Mitigating Predictive Multiplicity in Gradient Boosting
First systematic analysis of the Rashomon effect in gradient boosting with novel techniques to explore model sets and reduce predictive multiplicity for fairer model selection.
H. Hsu, I. Brugere, S. Sharma, F. Lecue, C.F. Chen
NeurIPS 2024
Comparing Apples to Oranges: Learning Similarity Functions for Data Produced by Different Distributions
Efficient sampling framework learning cross-group similarity functions with limited expert feedback, enabling fair comparisons of data from different demographic distributions.
L. Tsepenekas, I. Brugere, F. Lecue, D. Magazzeni
NeurIPS 2023
Additional Publications (2015–2026)
Revisiting ML Training under Fully Homomorphic Encryption: Convergence Guarantees, Differential Privacy, and Efficient Algorithms
Y. Zhou, M. Liang, I. Brugere, D. Dervovic, Y. Guo, A. Polychroniadou, M. Wu, D. Dachman-Soled
In submission
MAFE: Enabling Equitable Algorithm Design in Multi-Agent Multi-Stage Decision-Making Systems
Z.M. Lazri, A. Nakra, I. Brugere, D. Dervovic, A. Polychroniadou, F. Huang, D. Dachman-Soled, M. Wu
In submission
How Reliable are Confidence Estimators for Large Reasoning Models? A Systematic Benchmark on High-Stakes Domains
R. Khanmohammadi, E. Miahi, S. Kaur, C. Smiley, I. Brugere, K.S. Thind, M.M. Ghassemi
EACL 2026
The Unseen Threat: Residual Knowledge in Machine Unlearning under Perturbed Samples
H. Hsu, P. Niroula, Z. He, I. Brugere, F. Lecue, C.F. Chen
NeurIPS 2025
Balancing Fairness and Accuracy in Data-Restricted Binary Classification
Z. McBride Lazri, D. Dervovic, A. Polychroniadou, I. Brugere, D. Dachman-Soled, M. Wu
ACM TKDD
Investigating the Temporal Association of Biomedical Research on Small Business Funding: A Bibliometric and Data Analytic Approach
R. Khanmohammadi, S. Kaur, C.H. Smiley, T. Alhanai, I. Brugere, A. Nourbakhsh, M.M. Ghassemi
IEEE TCSS
A Canonical Data Transformation for Achieving Inter- and Within-group Fairness
Z. McBride Lazri, I. Brugere, X. Tian, D. Dachman-Soled, A. Polychroniadou, D. Dervovic, M. Wu
IEEE TIFS
Bounding the Accuracy Loss for Graphical Model Based Synthetic Data Generation in Privacy-Preserving Machine Learning
Y. Zhou, I. Brugere, D. Dachman-Soled, D. Dervovic, M. Liang, A. Polychroniadou, M. Wu
ICML 2023
Hyper-parameter Tuning for Fair Classification without Sensitive Attribute Access
A.K. Veldanda, I. Brugere, S. Dutta, A. Mishler, S. Garg
TMLR
Fairness via In-Processing in the Over-parameterized Regime: A Cautionary Tale with MinDiff Loss
A.K. Veldanda, I. Brugere, J. Chen, S. Dutta, A. Mishler, S. Garg
TMLR
Parameterized Explanations for Investor/Company Matching
S. Kaur, I. Brugere, A. Stefanucci, A. Nourbakhsh, S. Shah, M. Veloso
ICAIF’21 Workshop on Explainable AI in Finance
GAEA: Graph Augmentation for Equitable Access via Reinforcement Learning
G.S. Ramachandran, I. Brugere, L.R. Varshney, C. Xiong
AAAI AIES 2021
Evaluation of crowdsourced mortality prediction models as a framework for assessing artificial intelligence in medicine
T. Bergquist, T. Schaffter, Y. Yan, T. Yu, I. Brugere et al.
Journal of the American Medical Informatics Association
A continuously benchmarked and crowdsourced challenge for rapid development and evaluation of models to predict COVID-19 diagnosis and hospitalization
Y. Yan, T. Schaffter, T. Bergquist, T. Yu, J. Prosser, Z. Aydin, A. Jabeer, I. Brugere, et al.
JAMA Network Open
Network Structure Inference: Methodology and Applications
I. Brugere
Ph.D. Thesis
Network Structure Inference, A Survey: Motivations, Methods, and Applications
I. Brugere, B. Gallagher, T. Y. Berger-Wolf
ACM Computing Surveys
Network model selection with task-focused minimum description length
I. Brugere, T.Y. Berger-Wolf
WWW’18: BigNet Workshop
Coordination Event Detection and Initiator Identification in Time Series Data
C. Amornbunchornvej, I. Brugere, A. Strandburg-Peshkin, D. Farine, M.C. Crofoot, T.Y. Berger-Wolf
ACM TKDD
Evaluating Social Networks Using Task-Focused Network Inference
I. Brugere, C. Kanich, T.Y. Berger-Wolf
KDD’17: Workshop on Mining and Learning in Graphs
Both Nearest Neighbours and Long-term Affiliates Predict Individual Locations During Collective Movement in Wild Baboons
D. Farine, A. Strandburg-Peshkin, T.Y. Berger-Wolf, B. Ziebart, I. Brugere, J. Li, M. Crofoot
Nature Scientific Reports
Social Information Improves Location Prediction in the Wild
J. Li, I. Brugere, B. Ziebart, T. Y. Berger-Wolf, M. Crofoot, D. Farine
AAAI’15: Workshop on Trajectory-based Behaviour Analytics
Selected Patents
System and method for generating constrained loan pricing
Automated loan pricing system incorporating regulatory constraints and fairness requirements while optimizing financial objectives.
I. Brugere, M. Hosking, S. Sharma, F. Lecue, Y. Tan, J. Stettler, H. Zhao, P. Glover, D. Kapadia, G. Ciraulo, D. Bollum, D. Magazzeni, L.C. Liang
US Patent App. 18/397,698 · 2025
System and method for grounding outputs in tabular generative artificial intelligence
Method ensuring LLM-generated table analysis outputs are verifiable and traceable to source data for auditability in regulated applications.
I. Brugere, S. Kariyappa, S. Sharma, F. Lecue, G. Nguyen
US Patent 12,436,935 · 2025
System and method for graph-based resource allocation using neural networks
Neural network approach for optimizing resource distribution across graph-structured systems, enabling equitable allocation in networked environments.
G.S. Ramachandran, I. Brugere, L. Varshney, C. Xiong
US Patent 12,165,053 · 2024
Method and system for improving model fairness by using explainability techniques
Framework leveraging model explainability methods to identify and mitigate fairness issues by analyzing feature contributions across demographic groups.
I. Brugere, D. Magazzeni, N. Marchesotti, D. Heike, F. Zhao, E. Wang, H. Shu, M. Gabriel, M. Veloso, C. Tilli, S. Dutta, B. Mallik, Ade Onigbanjo
US Patent App. 17/968,220 · 2024

Service & Community

Workshop Organization (2016–2023)

  • NLP and Network Analysis in Financial ApplicationsACM ICAIF’23
  • PhD ForumIEEE ICDM’19
  • NetInf’17: First Workshop on Inferring Networks from Non-Network DataSIAM SDM’17
  • Inferring Networks from Non-Network DataSIAM AM’16

PC Member / Reviewer (2018–Present)

  • Conferences: AAAI, CIKM, FAccT, ICDM, ICLR, IJCAI, KDD, NeurIPS, PAKDD, SDM, TheWebConf, WSDM
  • Journals: ACM CSUR, IEEE TKDE, ACM TKDD, KAIS

Tutorial

  • Modeling Data with Networks + Network Embedding: Problems, Methodologies and Frontiers — I. Brugere, B. Perozzi, P. Cui, W. Zhu, J. Pei, T.Y. Berger-WolfKDD 2018

Other Service (2014–2020)

  • ACM Tapia Celebration of Diversity in Computing — Plenary Speaker2020
  • ACM Tapia Celebration of Diversity in Computing — Accessibility Committee2020
  • Bloomberg Data For Good Exchange Program Committee
  • Google Lime campus ambassador
  • University of Washington–AccessSTEM volunteer
  • ACM SIGKDD Broadening Participation in Data Mining Coordinator, Mentoring Co-Chair2014, 2016, 2017

Teaching

  • Teaching Assistant: Computer Algorithms I (Senior-level), University of Illinois at Chicago2017

Technical Skills

Languages

PythonJuliaScala

AI / ML

PyTorchscikit-learnXGBoostGymnasiumLM Studio

Graph Learning

DGLPyTorch GeometricNetworkX

Research Interests

Perturbation RobustnessMachine UnlearningConfidence EstimationFairnessGraph LearningPrivacy-Preserving ML
Ivan Brugere
Lead AI Research Scientist  ·  Chicago, IL
Get in touch Download CV
ivan@ivanbrugere.com  ·  github.com/ivanbrugere