Work
I'm an engineer/researcher at Johns Hopkins APL. In my first few months, I built research infra to study how different mixtures of simulated/real data affect self-, semi-, and fully-supervised algorithms for image recognition. Currently, I'm working on
- using neural fields to speed up electromagnetic scattering simulations
- building tool/retrieval augmented LMs ("agents") that can infer + suggest tactics in real-time strategy games
- evaluating adversarial attack + defense mechanisms for image recognition systems
RankCSE
While researching text embeddings, I implemented RankCSE,
a contrastive learning algorithm that models the fine-grained semantic similarity between texts.
The pretrained models are open source, and small enough to fit inside your local RAG application. They have been downloaded 27600+ times to date.
Code
- A basic recursive ray tracer
- Computing persistent homology for small 2D V-R complexes
Publications
| 1. | Composition-contrastive learning for Sentence Embeddings
with Ruihong Huang Association for Computational Linguistics, 2023 arXiv, GitHub |
Education
The Johns Hopkins University
MS, Electrical and Computer Engineering
2024-
Texas A&M University
BS, Computer Science
Minor in Math, Neuroscience
Craig and Galen Brown Engineering Honors
2019-2023