Semester and Thesis Projects

The ETH AI Center offers a wide range of semester and thesis projects for students at ETH Zurich, as well as other universities. Please see the list below for projects that are currently available.

How do you publish a thesis or semester project with the ETH AI Center?

Academia: If you are affiliated with the ETH AI Center as faculty member, post- or doctoral fellow, please add the following affiliation to your Sirop account. If you tag your thesis project with this affiliation, it should appear in the list below.
- 'ETH Competence Center - ETH AI Center (ETHZ)'
Industry: If you represent a company that has a corporate partnership with the ETH AI Center, please contact .

Need help?

We provide a external page template for thesis projects accouncements that you can use for your upcoming project & upload to external page SiROP.
Is your thesis project still missing from the list below? We are constantly adding new thesis projects that are available within the ETH AI Center.
Are you a student? Check out our Semester and Thesis projects below!

Master Thesis on AI agents for oncology

We are looking for a motivated Master thesis student to develop AI agent systems that support Molecular Tumor Boards, building on our NeurIPS paper (https://openreview.net/pdf?id=anzoPBV4jI) introducing MTBBench, a multimodal clinical decision-making benchmark in oncology.

Keywords

Agents, Oncology, Health, Language models, RL

Labels

Master Thesis , ETH Zurich (ETHZ)

PLEASE LOG IN TO SEE DESCRIPTION

More information

Open this project...

Published since: 2025-11-14 , Earliest start: 2025-11-16

Applications limited to ETH Zurich

Organization ETH Competence Center - ETH AI Center

Hosts Moor Michael

Topics Information, Computing and Communication Sciences

Synthetic Data Generation for Automated Speech Recognition for Impaired Speech

Automatic Speech Recognition (ASR) for individuals with impaired speech is severely hampered by data scarcity. This project addresses this problem by developing a personalized Text-to- Dysarthric-Speech (TTDS) model to serve as an advanced data augmentation method. Unlike assistive technologies that aim to correct speech impairments, the primary goal here is to faithfully clone a speaker’s unique impaired speech patterns. Using state-of-the-art generative audio models (e.g., VITS), the system will learn to generate synthetic yet realistically impaired speech data from very few recordings. A key innovation will be to leverage phoneme uncertainty analyses from prior work to guide the synthesis process, enabling the targeted generation of more realistic phonetic deviations. This project is designed for a highly motivated, independent individual ready to take ownership of a challenging research topic.

Keywords

Automated Speech Recognition (ASR), Data Synthesis, Personalization, Scarce Data Availablity

Labels

Master Thesis

Description

Goal

Contact Details

More information

Open this project...

Published since: 2025-11-03 , Earliest start: 2025-11-18 , Latest end: 2027-02-03

Organization ETH Competence Center - ETH AI Center

Hosts Böhringer Roman

Topics Information, Computing and Communication Sciences

MSc Thesis and Short Project - Developing an XR Platform for Demonstrating Human Perceptual Adaptation and Brain Plasticity

Human perception actively adapts to continuous sensory input, allowing the brain to recalibrate and maintain accurate representations of the world. Optical illusions, phantom sensations, and visuomotor distortions provide unique insights into this adaptability. At the Life Science Learning Centre (UZH and ETH), such e􀆯ects are currently demonstrated using physical setups. This Master’s short project will evaluate how Virtual Reality (VR) or Augmented Reality (AR) technologies could modernize and extend these neuroscience demonstrations. During this initial phase, you will explore commercially available systems and gain an understanding of the neuroscience behind perceptual illusions and adaptation. Once a feasible approach is identified, the prototype will be developed into a full implementation as part of a Master’s Thesis. This will include designing a behavioral experiment in which the XR solution is tested and evaluated with human volunteers. The overall goal is to enable users to experience visual and proprioceptive illusions interactively using immersive technology, ultimately contributing to educational and research applications in neuroscience and psychology.

Labels

Semester Project , Bachelor Thesis , Master Thesis , ETH Zurich (ETHZ)

Contact Details

More information

Open this project...

Published since: 2025-11-03 , Earliest start: 2025-11-03 , Latest end: 2026-04-03

Organization ETH Competence Center - ETH AI Center

Hosts Böhringer Roman

Topics Engineering and Technology , Behavioural and Cognitive Sciences

Personalizing Automatic Speech Recognition Models for Non-normative Speech using MoE

Automatic Speech Recognition (ASR) for individuals with impaired speech remains a significant challenge due to extreme data scarcity and high acoustic variability. This project builds upon a successful, data-efficient Bayesian personalization framework (Variational Inference Low-Rank Adaptation, VI-LoRA), coming from our team. This project aims to further increase personalised ASR performance by exploring the implementation of Hydra VI-LoRA. Hydra VI- LoRA, is a novel Mixture-of-Experts (MoE) architecture capable of learning specialized adapters for different speakers or phonetic challenges within a single model. This project is designed for a highly motivated and independent student eager to take ownership of a cutting-edge research topic.

Keywords

Automated Speech Recognition (ASR), Mixture of Experts, Personalization, data scarcity

Labels

Master Thesis

Description

Goal

Contact Details

More information

Open this project...

Published since: 2025-11-03 , Earliest start: 2025-11-15 , Latest end: 2027-04-01

Organization ETH Competence Center - ETH AI Center

Hosts Böhringer Roman

Topics Information, Computing and Communication Sciences

Toward Human-Like Perception with Multisensory AR Glasses

How can we give machines a perceptual experience closer to that of humans — fast, continuous, and adaptive? At the Sensors Group at the Institute of Neuroinformatics, we explore bio-inspired sensing and computation to bring human-like perception to real-world devices. In this project, you will work with state-of-the-art multimodal sensing devices, including AR glasses (RGB cameras, IMU, audio) and bio-inspired event cameras. Together, these sensors offer a unique opportunity to study how multiple sensory modalities can be fused to perceive and interpret the world in real time. Your research will advance real-time sensor fusion, embodied perception, and neural-inspired computation, forming the foundation for future research in AR scene understanding and augmented human-environment interaction.

Keywords

Multimodal Perception • Sensor Fusion • Augmented Reality • Event-Based Vision • Real-Time Systems

Labels

Semester Project , Bachelor Thesis , Master Thesis

PLEASE LOG IN TO SEE DESCRIPTION

More information

Open this project...

Published since: 2025-10-17 , Earliest start: 2025-10-20 , Latest end: 2026-02-28

Applications limited to ETH Zurich , University of Zurich

Organization ETH Competence Center - ETH AI Center

Hosts Liu Shih-Chii , Li Zixiao

Topics Information, Computing and Communication Sciences , Engineering and Technology

(For ETH students) Master’s Thesis in cooperation with Lufthansa Group (LHG): Improving Reliability and User-Centered Evaluation of Retrieval-Augmented Generation (RAG) Systems in the Aviation Airline TechOps Domain

Retrieval-augmented LLMs offer promising capabilities for supporting aircraft troubleshooting in airline technical operations (Airline TechOps). They can help engineers access large volumes of manuals, fault reports, and historical data quickly and efficiently. A previous thesis in cooperation with Lufthansa CityLine developed a framework to evaluate the performance of a RAG system for Airline TechOps use cases and identified issues influencing the system’s reliability. The feedback collected from engineers revealed persistent challenges: incorrect or missing references to technical manuals, hallucinated or irrelevant responses, inconsistent formatting, and too long or generic summaries. While some outputs were useful, user trust and adoption remain limited due to these issues. This thesis aims to address these shortcomings by designing evaluation and improvement methods that better align the RAG system with user needs, increasing its reliability and usability in the safety-critical contexts of aviation.

Keywords

LLM, RAG, Evaluation, NLP, user-centric

Labels

Master Thesis , ETH Zurich (ETHZ)

Description

**Master’s Thesis in cooperation with Lufthansa Group (LHG)** Improving Reliability and User-Centered Evaluation of Retrieval-Augmented Generation (RAG) Systems in the Aviation Airline TechOps Domain ** **Motivation:** Retrieval-augmented LLMs offer promising capabilities for supporting aircraft troubleshooting in airline technical operations (Airline TechOps). They can help engineers access large volumes of manuals, fault reports, and historical data quickly and efficiently. A previous thesis in cooperation with Lufthansa CityLine developed a framework to evaluate the performance of a RAG system for Airline TechOps use cases and identified issues influencing the system’s reliability. The feedback collected from engineers revealed persistent challenges: incorrect or missing references to technical manuals, hallucinated or irrelevant responses, inconsistent formatting, and too long or generic summaries. While some outputs were useful, user trust and adoption remain limited due to these issues. This thesis aims to address these shortcomings by designing evaluation and improvement methods that better align the RAG system with user needs, increasing its reliability and usability in the safety-critical contexts of aviation. **Objectives:** - How can hallucinations and incorrect references in RAG outputs be detected and reduced? - What evaluation methods best capture user-relevant quality dimensions (e.g., accuracy, grounding in resources, conciseness)? - To what extent can confidence scoring or fact-checking mechanisms improve trust in system outputs? - What strategies can enhance the system’s ability to reflect the structure and interrelations of technical documentation? - How can document retrieval be improved to better support relevance, completeness, and contextual grounding in RAG systems? **Methodology:** 1. Literature Review: Explore recent work on hallucination detection, factual consistency evaluation, and human-centered evaluation of (RAG-augmented) LLMs. 2. Error Analysis: Categorize failure cases based on feedback (reference errors, hallucinations, summarization issues, formatting). 3. Metric Development: Define or adapt automatic metrics that better capture the present concerns (e.g., reference correctness, conciseness). 4. System Enhancements: Implement and test approaches such as: - Reference validation - Confidence scoring or cross-checking outputs with multiple models. - Summarization post-processing for concise, actionable outputs. - Techniques to better represent and leverage dependencies between retrieved sources and generated responses. 5. User Study: Conduct a small-scale evaluation with users to measure improvements against baseline. 6. Regulatory Impact: Consider the regulatory dimension by aligning the framework with the EASA AI Roadmap and its latest Concept Paper for ML/AI applications. **Expected Outcomes:** - A refined evaluation framework linking technical metrics with user-perceived quality. - Prototype methods for reducing hallucinations and improving output consistency. - Improved modelling and utilisation of dependencies between retrieved and generated content. - Enhanced document retrieval mechanisms that improve relevance, completeness, and contextual grounding, supporting more accurate and actionable generation. - Empirical evidence of improved trustworthiness and usability of RAG outputs in aviation. - Recommendations or guidance to demonstrate trustworthiness for AI based operational tools in Airline TechOps domain (e.g. assurance blueprint). - A written master’s thesis and preferably an *ACL conference publication. **Required Skills:** - Experience with Python, LLM frameworks (HuggingFace, LangChain, vllm, etc.). - Interest in applied AI, safety-critical domains, and human-centered evaluation. - Prior knowledge of aviation maintenance processes is helpful but not required. **Supervision & Environment:** The thesis will be supervised by Dr. Mubashara Akhtar at the ETH AI Center/ETH Computer Science department and Marie-Suzanne Sarre from the industrial partner, Lufthansa Group.

Contact Details

More information

Open this project...

Published since: 2025-09-29 , Earliest start: 2025-10-15 , Latest end: 2026-06-30

Organization ETH Competence Center - ETH AI Center

Hosts Akhtar Mubashara

Topics Information, Computing and Communication Sciences

Engineering Wearable Platforms

This project focuses on the design, engineering, and prototyping of next-generation wearable technologies. The goal is to create microfluidic platforms that enable automatic sampling, reliable fluid control, and seamless integration of diverse sensing modalities.

Keywords

Wearables, materials, microfluidics

Labels

Internship , Master Thesis , Student Assistant / HiWi

PLEASE LOG IN TO SEE DESCRIPTION

More information

Open this project...

Published since: 2025-09-16 , Earliest start: 2025-09-21 , Latest end: 2026-09-01

Applications limited to Balgrist Campus , ETH Zurich , University of Zurich , EPFL - Ecole Polytechnique Fédérale de Lausanne , Empa , Zurich University of Applied Sciences

Organization ETH Competence Center - ETH AI Center

Hosts Dosnon Lucas

Topics Engineering and Technology

RA position: CoMind: Joint Interaction Understanding through Intent Prediction

Understanding human intent and anticipating future actions is crucial for enabling seamless human-robot interaction in real-world tasks. Intention prediction not only facilitates smoother collaboration with robots but also represents a fundamental challenge in the development of intelligent systems. In this project, we aim to collect a dedicated interaction dataset using the Meta Aria glasses and use the collected data to develop multimodal learning models that are capable of predicting human intent and anticipating future behaviors across a variety of tasks.

Keywords

egocentric vision, intent prediction, anticipation, multimodal learning

Labels

Semester Project , Master Thesis , Student Assistant / HiWi , ETH Zurich (ETHZ)

PLEASE LOG IN TO SEE DESCRIPTION

More information

Open this project...

Published since: 2025-07-14 , Earliest start: 2025-07-14

Organization Computer Vision and Geometry Group

Hosts Wang Xi , Kaufmann Manuel , Chen Jiaqi , Gavryushin Alexey

Topics Information, Computing and Communication Sciences