Multimodal Foundational Models For Neuroscience

Scholar Nexus

Paris

Sur place

EUR 16 000 - 22 000

Plein temps

14 jours+

Recevez plus de réponses des employeurs

Envoyez un CV adapté au poste en quelques minutes.

Avantages offerts par ce poste

Fully funded position

Tuition covered €400/year

Résumé du poste

Télécom Paris, part of Institut Polytechnique de Paris, invites applications for a PhD project on multimodal foundational models for neuroscience. The position is on-site in Paris and fully funded, with a focus on integrating EEG, fMRI, clinical records, and genomic data.

The research emphasizes knowledge-grounded learning using disease-specific graphs and EEG ontologies, enabling interpretable models. Strong computational and ML skills, especially PyTorch, are required, with excellent English.

Qualifications

Master’s degree in computer science, neuroscience, or related field with strong ML background.
Proficiency in English (advanced).
Experience with EEG data processing and multimodal data analysis is preferred.

Responsabilités

Develop multimodal foundational models integrating EEG, MRI, clinical, and genomic data.
Incorporate knowledge graphs and ontologies to ground representations.
Collaborate with large cohorts and ensure generalization across distributions.

Connaissances

Multimodal ML

Python / PyTorch

EEG data processing

MRI/fMRI analysis

Biomedical informatics

Formation

Master's degree in CS or Neuroscience

Outils

PyTorch

Description du poste

France/Télécom Paris (France)

On-site

Fully Funded

Multimodal Foundational Models For Neuroscience

Description

This PhD project at Télécom Paris, part of Institut Polytechnique de Paris, addresses a critical challenge in computational neuroscience: developing multimodal foundational models that can effectively integrate and analyze heterogeneous neurological data. The position sits at the intersection of machine learning, neuroscience, and biomedical informatics.

The research focuses on building foundational models that can process multiple data modalities including high temporal resolution EEG, functional/structural MRI (fMRI/sMRI), clinical records, and genomic profiles. Current approaches largely develop models for single modalities in isolation, missing the complementary information available across different data types. This project aims to overcome this limitation by creating integrated frameworks.

A key innovation of this work is the incorporation of structured biomedical knowledge. Rather than learning representations from data alone, the models will leverage disease-specific knowledge graphs (such as those for Alzheimer's disease or epilepsy) and EEG-specific ontologies. This knowledge-grounded approach aims to produce results that are not only accurate but also clinically interpretable and explainable.

The project builds on recent advances in foundational models for EEG and fMRI, including large brain models and foundation models for brain activity recordings. However, it goes beyond existing work by focusing on multimodal integration and knowledge incorporation. The successful candidate will work with large-scale cohorts and develop models capable of generalizing across different data distributions and applications.

The position requires strong computational skills, particularly in machine learning and deep learning frameworks like PyTorch. Candidates should have experience with EEG data processing and be comfortable working with complex, heterogeneous datasets. The project offers the opportunity to contribute to cutting-edge research at the intersection of artificial intelligence and neuroscience, with potential applications in understanding neurological mechanisms and disease.