Summary
Sonia Joseph is a visiting researcher and PhD candidate bridging neuroscience, computer science, and machine learning with eight years of hands-on experience in both academic and industry labs. She currently researches interpretable, physically plausible world models on Meta’s JEPA team while leading a 160+ member interpretability forum across FAIR, Reality Labs, and GenAI. Her work spans multimodal mechanistic interpretability—contributing to MATS cohorts and authoring the open-source Prisma library for vision interpretability and sparse autoencoders—and she co-founded Alexandria Labs, moving from CTO to strategic advisor after early-stage growth. With roots in computational neuroscience at Princeton and Janelia, she combines deep experimental analysis (calcium imaging, receptive field pipelines) with ML engineering for large models and RAG-style systems. Known for a playful AI “zoo” on GitHub and active community mentorship, she blends rigorous research with practical tooling that accelerates model understanding.
8 years of coding experience
3 years of employment as a software developer
Master's degree Computer Science, Master's degree Computer Science at McGill University
High School Diploma, High School Diploma at Acton-Boxborough Regional High School
Doctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at Mila - Quebec Artificial Intelligence Institute / McGill
Bachelor of Arts (B.A.) Neuroscience Computer Science Machine Learning Literature, Bachelor of Arts (B.A.) Neuroscience Computer Science Machine Learning Literature at Princeton University
English, Latin