LLMs and Text Processing

Target audience

Participants interested in developing practical skills in Large Language Models and Generative AI, with a focus on applications in competition and regulatory contexts. The course is particularly suited to participants who want to build and evaluate LLM-based solutions such as RAG document assistants using Python and real regulatory documents.

Intended Learning Outcomes

At the end of the course, participants will be able to:

1. Understand and control LLMs: explain the Transformer architecture, manipulate inference parameters (temperature, top-p) to control output, and distinguish between pre-training and fine-tuning.
2. Implement Generative AI solutions: use Python to interact with LLMs via both APIs (OpenAI or other) and local open-source libraries (Hugging Face).
3. Build RAG systems: architect and code a complete RAG pipeline, including document chunking, embedding creation, vector storage, and context-aware generation.
4. Evaluate and optimize: identify common failure points in Generative AI applications (hallucinations, retrieval failures) and apply advanced techniques (re-ranking, tool use/agents) to mitigate them.

Estimated Time for Completion

12 hours

Level

Level 6: Bachelor’s degree (first cycle)

Type of the assessment

Quizzes and project evaluation

Form of participation in the learning activity

• Asynchronous learning (videos, quizzes, exercises)
• live training workshop (exercise resolution and clarification)
• Project work in groups

Pre-requisites

• Solid understanding of Python (data structures, functions).
• Basic familiarity with concepts from math and linear algebra (vectors, matrices, basic probability).