Theses | Giordano d'Aloisio

📖 Open Theses

The following projects are available for students. If you are interested or want more details, please contact me.

Automating the Compression of Large Language Models of Code
Research and implement search‑based strategies to identify effective model compression pipelines that reduce model size while preserving performance. The work involves designing search procedures, evaluating compression candidates, and benchmarking tradeoffs between size, speed, and accuracy.

A.Y. 2025-2026
- Implementation and Evaluation of a Low‑Code Web Application for the Development of Fair Learning‑Based Systems
  - Candidate: Francesca Controguerra
  - Supervisor: Prof. Antinisca Di Marco – University of L’Aquila
- Optimizing Quality and Sustainibility of Code Generation Agents
  - Candidate: Kasra Sartaee
  - Supervisor: Prof. Federica Sarro – University College London

A.Y. 2025-2026
- Understanding and Mitigating Gender Bias in Stable Diffusion 3: An Analysis of Embedding-Level Intervention
  - Candidate: Asmita Anand
  - Supervisor: Prof. Federica Sarro – University College London
A.Y. 2024-2025
- The Moral Cost of Visual Bias: Intersectionality in AI Image Generation Across Religion, Ethnicity, and Gender
  - Candidate: Tabatha Adorno Correa
  - Supervisor: Dr. Donatella Donati – University of L’Aquila
- How Do Generative Models Draw a Software Engineer? Investigating and Mitigating Bias in Stable Diffusion Models
  - Candidate: Tosin Fadhaunsi
  - Supervisor: Prof. Federica Sarro – University College London
- Improving Energy Consumption of Text2Image Generation Models
  - Candidate: Jay Choy
  - Supervisor: Prof. Federica Sarro – University College London
- Towards Fair and Sustainable Machine Learning: A Multi-Objective Ensemble Optimization Approach
  - Candidate: Vayk Mathrani
  - Supervisor: Prof. Federica Sarro – University College London