Blake Bullwinkel

I'm a Researcher on the AI Red Team at Microsoft, where I develop attacks against generative AI systems to uncover safety and security vulnerabilities. By studying these failure modes, I hope to both mitigate applied AI risks in the short term and understand how to build more robust systems in the long term. My interests in safety and alignment began during my master's thesis research, which focused on incorporating physics-based constraints into neural networks. I was lucky to receive academic mentorship from Pavlos Protopapas and Weiwei Pan.

In my free time, I enjoy running and being outside. I love learning and sometimes work on a page of math notes to explain important topics in my own words. You can find my resume here and should feel free to reach out!

Research

Lessons From Red Teaming 100 Generative AI Products
B Bullwinkel et al.
Microsoft BlueHat, 2024
NeurIPS Workshop on Red Teaming GenAI, 2024
Blog Arxiv eBook Talk

Phi-3 Safety Post-Training: Aligning Language Models with a "Break-Fix" Cycle
B Bullwinkel et al.
Arxiv, 2024
Arxiv

PyRIT: A Framework for Security Risk Identification and Red Teaming in Generative AI Systems
B Bullwinkel et al.
Conference on Applied Machine Learning in Information Security (CAMLIS), 2024
Arxiv

Using Large Language Models for Humanitarian Frontline Negotiation: Opportunities and Considerations
Z Ma, S Su, N Zhao, L Bieske, B Bullwinkel, Y Zhang, S Yang, Z Luo, S Li, G Liao, B Wang, J Gao, Z Wen, C Bruderlein, W Pan
ICML Workshop on the Next Generation of AI Safety (NextGenAISafety), 2024
Arxiv Poster

Transfer Learning with Physics-Informed Neural Networks for Efficient Simulation of Branched Flows
R Pellegrin, B Bullwinkel, M Mattheakis, P Protopapas
NeurIPS Workshop on Machine Learning and the Physical Sciences, 2022
Arxiv Poster

DEQGAN: Learning the Loss Function for PINNs with Generative Adversarial Networks
B Bullwinkel, D Randle, P Protopapas, D Sondak
ICML Workshop on AI for Science (AI4Science), 2022
Arxiv Poster

Evaluating the Fairness Impact of Differentially Private Synthetic Data
B Bullwinkel, K Grabarz, L Ke, S Gong, C Tanner, J Allen
ICML Workshop on Theory and Practice of Differential Privacy (TPDP), 2022
Arxiv Poster

Projects

Azure/PyRIT
Active contributor to PyRIT, an open source framework that empowers security professionals and machine learning engineers to proactively find risks in their generative AI systems.
Repo

DEQGAN
Co-created a Python package that implements DEQGAN, an unsupervised generative adversarial network method for solving ordinary and partial differential equations.
Repo

Marble Groceries
Led a team of engineers to develop an iOS app that helps users understand the environmental impact of their grocery purchases by scanning product barcodes.
App Store

Linearized Neural Nets for Transfer Learning with GPs
Implemented the method for transfer learning proposed in Fast Adaptation with Linearized Neural Networks, a 2021 paper by Maddox et al., in TensorFlow and performed experiments to test its practical utility.
Notebook Code

Classifying the Sounds of NYC
Trained and tuned a variety of models to classify audio clips recorded around New York City from the UrbanSound8k dataset into ten different classes.
Report Notebook

Modeling ASA Section Membership
Constructed binary response generalized linear models to predict whether or not members of the American Statistical Association belonged to at least one section.
Report Code

Forecasting and Classifying Mice Microbiomes
Used deep learning to forecast qPCR time series and classify mouse microbiomes into healthy and infected groups based on data provided by researchers at Brigham and Women's Hospital.
Notebook

Woof Woof! Computer Vision & NLP App for Austin Pets Alive
Attended workshops at the 2021 IACS ComputeFest to build a web app that allows users to "chat" with and search for visually similar dogs in the Austin Pets Alive animal shelter by leveraging backend NLP and computer vision models.
Code

Wildfire Risk Prediction & Response Optimization
Trained tree-based classification models on historical wildfire and weather data to predict the fire risk for a given county and month in California and used mixed-integer programming to determine the optimal assignment of limited firefighters across the state, based on total cost.
Report Code

Predicting the Outcome of the 2020 Election
Built k-NN and regularized logistic regression models to predict the outcomes of the 2020 presidential and congressional elections using historical, polling, and fundamentals data.
Report Code

DreamDiff Python Package
Worked in a team of three to develop a Python package that implements forward-mode automatic differentiation (AD), root-finding, optimization, and quadratic spline interpolation.
Repo PyPI

Analysis of Wildfires, Air Quality, and Public Health
Conducted time series analysis in R to link spikes in PM2.5 concentration to specific wildfire events in California and used major axis regression to explore correlations between air quality and public health outcomes.
Slides Code

Predicting Agricultural Crop Quality
Built a variety of regression models to predict the quality of crops based on agricultural and weather data.
Slides Code

Modeling Electricity Consumption in the US
Built linear regression models to predict household electricity consumption in the US from various residential characteristics.
Report Code

Early Epidemiological Model Parameters for COVID-19
Modeled early-stage COVID-19 case data in mainland China using systems of ordinary differential equations.
Slides Code