Master Thesis: Machine Learning classification in Integrated Sensing and Communication

Join our Team

About this opportunity:

Modern radio networks can be used as radar-like sensors to detect and classify moving objects — a capability broadly referred to as Integrated Sensing and Communication (ISAC). Our team builds the machine-learning models behind this capability.  

As we deploy the models to different locations and extend it to more object classes, two operational problems dominate our research agenda: 

• Cross-location generalisation: models trained on data from one location do not always transfer to another. Every new location currently requires labelled data-to attune for the new location — a slow and expensive step that gates deployment. 


• Multi-class extension: adding new object classes without degrading performance on existing ones. It involves architectural trade-offs that are not obvious. 
  
  

We are offering three related master's thesis opportunities on these problems, all rooted in production machine learning. Please indicate in your application which direction you are most interested in. The exact scope of each project will be refined together with the student based on their interests, background, and the state of our data and models when the project starts.  

What you will do:

• Direction 1: Cross-location Generalisation of a spectral image Classifier 

Benchmark modern domain-generalisation methods for image data. Implement source-free adaptation methods that adjust the model at deployment using only unlabelled data from the new site. 

• Direction 2: Cross-location Generalisation of a time-series Classifier 

Benchmark modern domain-generalisation methods for time-series data. Implement source-free adaptation methods that adjust the model at deployment using only unlabelled data from the new site. 

• Direction 3: Modular Classifier Design for Multi-Class Object Classification 

Benchmark methods for multi-class classification models such as Hierarchical tree, Mixture-of-experts and flat softmax. Evaluate how well they perform when adding additional classes and seeing new locations.

The skills you bring:

• You are pursuing a master's degree in computer science, machine learning, applied mathematics, or a related field. 


• Good understanding of modern machine learning, particularly deep learning. 


• Strong programming skills, familiarity with deep learning frameworks, and an interest in experimental, research-oriented work.  


• You are analytical, curious, and able to work independently while communicating your results clearly.