Visiting Researcher at UCSF (Rauschecker–Sugrue Lab) working on machine learning for clinical neuroimaging and signal processing.
Current work focuses on longitudinal trajectory prediction using the ABCD Study dataset and DBS connectivity fingerprints to validate the destination network convergence hypothesis. Particularly interested in using self-supervised learning to improve the robustness of neural decoding under physical hardware constraints.
Dual background in Materials Engineering (UTC) and Computer Science (École Polytechnique, MX25).
Weak Supervision for White Matter Lesion Detection Swin UNETR pipeline for leukoaraiosis detection across 11,868 adolescents (ABCD Study) — no voxel-level annotations, extreme class imbalance, 21-site heterogeneity.
DBS Connectivity Fingerprint Analysis SIFT2-weighted tractography fingerprints from 116 stimulation configurations in 5 DBS patients. Tests the destination network convergence hypothesis with 7 ML classifiers.
Multi-Scale Temporal Masking for Self-Supervised fMRI Learning Extensions to Brain-JEPA: hierarchical masking at three temporal scales and VICReg covariance regularization to prevent representational collapse.