This course introduces the fundamental principles of large language models (LLMs) and their applications in bioengineering and biomedical research. Students will learn how LLMs process and generate text, how they can be adapted to scientific and medical domains, and how to use them responsibly. Through examples and short hands-on exercises, participants will explore how LLMs can support literature analysis, knowledge extraction, and decision support in healthcare and life sciences. Learning objectives: Understand the basic architecture and functioning of LLMs (tokenization transformers, attention) Learn prompt engineering and retrieval-augmented generation (RAG) techniques Apply LLMs to biomedical text analysis and summarization tasks Discuss ethical, reliability, and privacy aspects of using LLMs in medicine and biology.

Guest Lecture: Prof. Dr. Umberto Michelucci is an expert in artificial intelligence and machine learning with a focus on scientific and medical applications. He teaches and conducts research at the Lucerne University of Applied Sciences and Arts (HSLU) and has extensive experience in applying data-driven and deep learning methods to biomedical and physical systems. His professional background includes academic research, industrial collaboration, and AI product development. 1. Introduction to Generative AI and LLMs Overview of generative AI; differences between traditional ML and LLMs; main applications in bioengineering and healthcare. 2. Fundamentals of LLMs Language representation, tokenization, embeddings; transformer architecture and attention mechanisms; overview of major models (BERT, GPT, Llama). 3. Working with Biomedical Text Data Sources of biomedical and clinical text (PubMed, ontologies, patient records); preprocessing, anonymization, and contextual understanding. 4. Prompt Engineering and Interaction Principles of effective prompting; few-shot and zero-shot learning; hands-on practice with open LLM interfaces. 5. Adapting and Extending LLMs Introduction to fine-tuning and retrieval-augmented generation (RAG); practical example using biomedical literature search. 6. Evaluation, Bias, and Ethics Understanding hallucinations, factual accuracy, and model bias; responsible use of LLMs in medical and scientific contexts. 7. Hands-On Mini-Project Students build a small application such as a biomedical Q&A assistant, literature summarizer, or terminology extractor.

In presenza

Presentazione orale

P.D.2-2 - Aprile