Sigma Nova is recruiting M2/PhD research interns to work on foundation models for brain signals, with a focus on robust generalization under strong data shift (cross-subject, cross-session). Our team has won the silver medal on 2025 NeurIPS EEG Challenge. A few key points differentiate this opportunity:

• You’ll work primarily with EEG, with opportunities to connect to other modalities (MEG/iEEG/fMRI)
• You’ll do publication-oriented foundation model research at the intersection of machine learning and neuroscience
• You’ll have access to large-scale compute, and large amounts of data, including in-house data acquired through partnerships

Artificial Intelligence (AI) models are frequently deployed in contexts different from the well-behaved, clean datasets they were trained on. In the literature, this phenomenon is known as distribution shift, a case where the underlying probability distribution from test data differs from the training data.

The distribution shift problem occurs frequently in Brain Computer Interfaces (BCIs), where training data is composed of a heterogeneous ensemble of readings from different subjects, and must generalize to unseen subjects. This setting poses the well-known cross-subject variability in EEG data, that is, models have to cope with the inherent heterogeneity of the training data and test data.

This internship seeks to study the impact of cross-subject induced distribution shifts in EEG foundation models. The intern will select one of the possible research lines to follow,

1. New pre-training strategies
2. New fine-tuning strategies
3. New in-context learning strategy
4. New data generation strategies tailored for EEG

In summary, this internship is at the intersection of domain adaptation, model weight space learning and interpretability of foundational models, intersecting every step of the large-scale model life-cycle pipeline.