Hi, I'm Baimam Boukar 👋
I am a researcher working on mechanistic interpretability and AI alignment. I study why current interpretability methods fail to generalize, and build more reliable tools to understand and control large language models. I recently completed my master's in Applied Machine Learning at Carnegie Mellon University, and I am currently a Research Assistant at Jinesis AI Lab, University of Toronto, working on the generalizability of mechanistic interpretability techniques.
Jul 9, 2026 Intellibra won Cameroon's Social Entrepreneurship Prize (1st Place)
Research Focus
Mechanistic Interpretability
Understanding internal model representations, and studying why interpretability methods often fail to generalize.
Deceptive Alignment and Situational Awareness
Detecting and analyzing deception, misalignment, and emergent behaviors inside large language models.
White-box Control
Building reliable activation-level tools — probes, steering, and patching — to monitor and control model behavior.
Science of Evals
Evaluation methodologies for emergent capabilities in LLMs, with norm-matched controls.
Selected Work
View AllFeatured Projects
Selected Publications
Phoenix: Safe End to End Codebase Refactoring via Multi-Agent LLMs
Baimam Boukar · with K. Koech, M. Adam, J. Barros
International Conference on Responsible AI (ICRAI) 2026•Published
Zero-Shot Neural Priors for Generalizable Cross-Subject and Cross-Task EEG Decoding
Baimam Boukar · with B. Fonya, N. Tagha Ghogomu, P. Nyaboe
Medical Imaging with Deep Learning (MIDL) 2026•Submitted
Retrieval with Multiple Query Vectors through Anomalous Pattern Detection
Baimam Boukar · with A. Nken, M. Rateike, C. Cintas
AAAI Conference on Artificial Intelligence 2026•Published