We are looking for a motivated Master’s student to investigate why modern audio-language models struggle to use clinical information to improve speech recognition for people with speech impairments as part of a Master’s thesis project. The project will focus on understanding the internal mechanisms of these models, identifying why additional context can negatively affect transcription accuracy, and exploring potential solutions. The work involves machine learning, mechanistic interpretability, and speech recognition, with opportunities to contribute to a scientific publication. The thesis will be conducted at the Institute of Neuroinformatics, ETH Zurich, in close collaboration with researchers at Cornell University.

We are looking for a motivated Master’s student to investigate how automatic speech recognition models can continuously adapt to the speech of individuals with speech impairments while minimizing the effort required from users as part of a Master’s thesis project. The project will focus on active learning, uncertainty-guided data selection, synthetic training data, and automated feedback to improve personalized speech recognition over time. The thesis will be conducted at the Institute of Neuroinformatics, ETH Zurich, as part of our Adapt4Me research project, with the opportunity to contribute to a scientific publication.

Learning and memory depend on changes in the strength of connections between neurons. In cortical pyramidal neurons, many long-range inputs arrive at apical dendrites, far from the cell body. How these distant signals contribute to synaptic plasticity remains an important open question. Strong dendritic depolarisations, including plateau potentials, may provide a mechanism linking distant synaptic input to neuronal output. Their effects may depend on input timing and inhibitory control by local interneurons. This thesis is part of our research on single-cell computations in the medial prefrontal cortex (mPFC). In this project you will first develop hands-on expertise in slice electrophysiology and related experimental methods, then pursue one clearly defined subproject depending on technical progress and the status of ongoing experiments

The main objective of this Master's thesis is to establish and validate a new automated visual-motor learning paradigm in mice. This involves developing a reliable home-cage training system in which animals can voluntarily participate in a visual-motor task. Based on that the student will be implementing a curriculum-based training paradigm, for which task difficulty and learning progression is gradually increased according to each animal's progress. The student will contribute to integrating sensors, developing automated visual stimulus and reward control, and establishing reliable methods for monitoring and recording behavioural performance. Once the system is established, the student will run a first cohort of mice to evaluate the training procedure, investigate learning dynamics and identify potential improvements to the experimental setup.