Summary
Aaron Mueller is an Assistant Professor and NLP researcher with a decade of experience probing what large neural models know and how to make them more robust, interpretable, and editable. He bridges academia, industry, and defense labs—having held research roles at Johns Hopkins, Meta, AWS, Raytheon, and Northeastern—bringing practical experience in multilingual MT, few-shot intent classification, and model editing. His work focuses on targeted model editing, concept erasure, and revealing decision-making mechanisms in deep networks to improve reliable generalization across languages. Aaron’s research combines rigorous evaluation (syntactic and multilingual probing) with applied improvements to production-style systems, such as speed-accuracy trade-offs in MT and BLEU gains for low-resource languages. Based in Boston, he brings a linguistics and computer-science background (BS dual degree, MS and PhD from Johns Hopkins) that informs both theoretical insight and hands-on system building. Colleagues note his knack for turning interpretability findings into concrete editing methods that change model behavior without wholesale retraining.
10 years of coding experience
8 years of employment as a software developer
Johns Hopkins University
Visting Academic, Data Science, Visting Academic, Data Science at New York University
B.S., Linguistics; B.S., Computer Science, 4.0, B.S., Linguistics; B.S., Computer Science, 4.0 at University of Kentucky
English, french (québec)