About me
I’m a statistician & AI / Machine learning researcher with a background in Social Science.
I think about how to use science and technology to improve society.
See my experience below for more info on my work and the Projects page for examples.
I also offer consulting services for statistical modeling, AI and machine learning methodology, computer vision, multi-agent reinforcement learning & agent-based modeling, and behavioral economics.
Experience
Post-Doc @ Lawrence Livermore Nat’L Lab: Machine Learning Group, Computational ENgineering Division
January 2024 - Present
IBM, AI for social Good research fellow
May 2021 - August 2021
Over the summer, we worked with organizations from different areas of the social sector and devised a common, reusable solution framework to address their needs. The team and I developed a novel restless multi-armed bandit algorithm to support the decision making of Breaking Ground in homeless outreach, Change Machine in financial coaching for socioeconomically disadvantaged people, and Leket in gleaning unharvested food.
Pacific Northwest National Laboratory (PNNL), AI & Machine learning Research intern
May 2019 - August 2019
Coded and tested 5 convolutional architectures based on UNet, ResNet, and SegNet for image segmentation with standard and custom loss functions. Obtained human-level performance for image labels. Reduced time spent labeling images by costly personnel from weeks to minutes. Manuscript published. Ported deep learning code from Keras to PyTorch. Wrote review on image segmentation with small sample sizes.
Managed 2 interns to write report on challenges in applying reinforcement learning to PNNL’s Airport Risk Assessment Model (ARAM) for automated parameter tuning. Proposed suite of agent-based models for testing ARAM performance. Proposed a hierarchical reinforcement learning (RL) approach to coordinating resource allocation among U.S. Customs & Border Patrol’s chemistry labs. Wrote review on adversarial RL techniques for optimizers used in recommender systems.
Presented an introduction to action-value based tabular RL for PNNL’s Summer Tech Talks.
Wrote review for applying hypergames in attacker-defender scenarios, agent-based models, and RL.
Duke Clinical Research Institute: AI Health Unit, Statistics Trainee & T32 Fellow
August 2018 - Present
Developing an unsupervised learning method for detecting and presenting anomalies in hearts based on unlabeled ultrasound videos. Manuscript in progress. Advisor: Dr. Ricardo Henao
Applied Bayes bootstrap with Thompson Sampling to Multi-Armed Bandits under correlated rewards to simulate effects of treatment “spillage" among clinical trial patients for online treatment assignment. Advisor: Dr. Alaatin Erkanli
–
NCSU Laboratory for analytic sciences, Graduate student researcher
August 2017 - December 2019
Developing theory based on multi-agent reinforcement learning to optimize large multi-agent systems, with applications in sociology and criminology. Manuscript in progress.
Managed small team to develop an open source Python-based wrapper package Open ABM for streamlining construction and analysis of agent-based models (first release in December 2019)
Wrote two agent-based model simulations in Python for applications in illicit network studies: one simulating a cartel, and the other petty crimes in a city. Book chapter published.
Interests
Cycling
AI for Social Good
Watercolor painting
Sci-Fi
Baked goods
Photographing my cats
Music production
Education
North Carolina State university, raleigh, nc
Ph.D., STATISTICS (Fall 2023)
M.S. STATISTICS
December 2020
University of Georgia, athens, ga
A.B./M.A. ECONOMICS,
May 2015
university of georgia, athens, GA
B.S. MATHEMATICS
May 2015