MASTER THESIS

MASTER THESIS

.NeuroRestore is a research and innovation center spanning EPFL, CHUV and UNIL that develops bioengineering strategies, including neurosurgical interventions, to restore neurological function. Building on more than a decade of preclinical work in rodent and non-human primate models, we integrate implantable neurotechnologies and novel therapies for paraplegia, tetraplegia, Parkinson's disease, stroke, and traumatic brain injury.

Layer-wise Representation Analysis of EEG/ECoG Foundation Models for Clinical Decoding

Foundation models are reshaping how we decode the brain, yet in clinical settings they are usually used as black boxes, with no principled way of choosing which of their internal representations to trust. .NeuroRestore is offering an internship or master's thesis position (4-6 months) to help change that: investigating how EEG foundation models represent information internally, and how to best extract and exploit these representations for clinical decoding.
The candidate's main duties include:

  • Running a large-scale analysis of state-of-the-art EEG foundation models across different downstream tasks and datasets, to develop a new methodology for using EEG foundation models in clinical settings

  • Building a pipeline to run computationally heavy analyses on a large GPU cluster

  • Training downstream decoders on top of existing EEG foundation models

  • Reading the most recent scientific literature about EEG Foundation Models

  • Contributing clean, documented, version-controlled code

The candidate will have:

  • Enrolled in an engineering, computer science, or data science master's program from an accredited institution

  • Strong computational and data analysis skills, with solid Python and PyTorch experience

  • Good knowledge of state-of-the-art machine learning techniques, including transformer architectures and self-supervised pre-training

  • Prior knowledge of EEG/ECoG processing is a plus but not necessary

  • Experience with Git, for collaborative development and code management

  • Strong aptitude for teamwork and problem-solving, excellent communication skills

  • Rapid learning capabilities, demonstrated independence and proactive approach in tasks

  • Fluent in English

Dates and Length:
Preferably starting in September or October 2026 for 4-6 months, but can be discussed.
Location:
The candidate will work at our .NeuroRestore offices at CHUV campus, Lausanne.
Contact:
Applications, including a CV and a short motivation letter (max 2 paragraphs), should be sent to martin.picek@epfl.ch.

Yue Yang Teo