Thesis Proposals

I am part of the Machine Learning Research Group, whose webpage collects thesis opportunities offered by all members of the group. I encourage students to also explore the proposals available there, as they provide a broader overview of the research topics currently pursued within the group and may help you identify the thesis project that best matches your interests.

Below are some thesis project ideas that I am available for supervision. These are research-oriented thesis projects and may lead to scientific publications, depending on the results achieved during the work. Of course, if you have a specific idea in mind, you can contact me via email to discuss it further.

Graph Generative Models for de Novo Drug Design - Master’s thesis, 6 months.

Study generative graph-based models for designing novel molecules from scratch. The project focuses on representing molecules as graphs and generating valid chemical structures. Possible goals include improving novelty, validity, and chemical usefulness. This topic is suitable for students interested in AI, chemistry, and drug discovery.

Graph Generative Models for Generating New Anticancer Molecules - Master’s thesis, 6 months.

Investigate graph generative models for creating new molecules with potential anticancer activity. The work may include molecular generation, filtering, and basic evaluation of candidates. The goal is to explore AI methods that can support early-stage cancer drug discovery. This topic is suitable for students interested in AI, chemistry, and drug discovery.

Molecule Representations in Graph Generative Models - Master’s thesis, 6 months.

Explore how different molecular representations affect graph generative models. The thesis may compare encodings, graph structures, or learned embeddings. The aim is to understand which representations help models generate better molecules. This project is suitable for students interested in machine learning foundations.

Molecule Toxicity Prediction - Master’s thesis, 6 months.

Develop machine learning models to predict whether a molecule may be toxic. The project may involve data preprocessing, feature extraction, and model evaluation. The goal is to support safer molecule selection in drug discovery pipelines. This topic is relevant for students interested in AI for health and chemistry.

Anticancer Molecules Activity Prediction - Master’s thesis, 6 months.

Build predictive models to estimate the biological activity of anticancer molecules. The work may include molecular descriptors, graph models, or deep learning methods. The objective is to identify compounds with promising anticancer potential. This project connects machine learning with pharmaceutical research.

Large Language Models for NER Task - Master’s thesis, 6 months.

Investigate the use of large language models for Named Entity Recognition (NER) tasks. Possible directions include entity extraction, domain adaptation, or improving robustness across different datasets. The aim is to explore how LLMs can enhance performance and flexibility in NER applications. This topic combines natural language processing with modern deep learning techniques.

Large Language Models for Blockchains - Master’s thesis, 6 months.

Investigate applications of large language models in the blockchain domain. Possible directions include smart contract analysis, blockchain data interpretation, or question answering. The aim is to explore how LLMs can support understanding and automation in this field. This topic combines natural language processing with blockchain technologies.

Large Language Models and Graph RAGs for Low-Compute Environments - Master’s thesis, 6 months.

Investigate the use of large language models combined with graph-based retrieval-augmented generation (RAG) in low-computation scenarios such as mobile devices. Possible directions include efficient graph construction, lightweight retrieval methods, or model optimization for limited hardware. The aim is to explore how LLMs can provide effective and scalable solutions under resource constraints. This topic combines natural language processing, graph-based methods, and efficient deep learning.