Fundamantels of Large Language Models

AIM

This course provides a comprehensive study of Large Language Models (LLMs), covering their architecture, underlying methodologies, operational principles, and practical considerations. Students will develop a deep understanding of transformer-based architectures, including attention mechanisms, tokenization, embeddings, encoder-decoder and decoder-only models, and the mathematical foundations that enable large-scale language understanding and generation. Students will learn how LLMs process information, generate outputs, reason over context, and interact with external knowledge sources. The course examines the complete LLM lifecycle, including data collection and preparation, pre-training, fine-tuning, instruction tuning, reinforcement learning from human feedback (RLHF), retrieval-augmented generation (RAG), prompt engineering, model evaluation, and deployment strategies. The course also includes critical topics such as efficiency, scalability, bias, safety, security, and responsible AI development. As a semester project, students will either develop an innovative LLM application for a real-world problem or conduct an in-depth research study on a contemporary topic in Large Language Models.

CONTENT

This course contains; Introduction to NLP and Language Models,Tokenization, Word Embeddings and Representational Learning,LLMs Basics,Attention Mechanism in Transformers,The Transformer Architecture,Transformer-Based Architectures and Models,LLM Training,Midterm Week,LLM Inferencing, Prompting, ICL, RAG, FT, Prompt Engineering ,Information Retrieval and Retrieval Augmented Generation - RAG, Parameter Efficient Fine Tuning, LMM Model Compression,Agentic AI and Agentic AI Frameworks,Model Context Protocol (MCP),Multi-Modal LLMs,Review and Presentations.

1. 1

   By the end of the course, students will be able to: 1. Explain the architecture and training pipeline of modern LLMs. 2. Implement core components of Transformer-based language models. 3. Fine-tune and evaluate pretrained LLMs for downstream tasks. 4. Understand alignment, instruction tuning, and human-feedback pipelines. 5. Analyze efficiency, scaling, and deployment constraints. 5. Students may develop and deploy LLM-based applications, critically review and reproduce recent research on Large Language Models (LLMs), or conduct original research on novel LLM architectures, applications, prompting techniques, fine-tuning strategies, evaluation methods, safety, or efficiency improvements.

WEEKLY PLAN

1. WEEK 1

   Introduction to NLP and Language Models

2. WEEK 2

   Tokenization, Word Embeddings and Representational Learning

3. WEEK 3

   LLMs Basics

4. WEEK 4

   Attention Mechanism in Transformers

5. WEEK 5

   The Transformer Architecture

6. WEEK 6

   Transformer-Based Architectures and Models

7. WEEK 7

   LLM Training

8. WEEK 8

   Midterm Week

9. WEEK 9

   LLM Inferencing, Prompting, ICL, RAG, FT, Prompt Engineering

10. WEEK 10

    Information Retrieval and Retrieval Augmented Generation - RAG

11. WEEK 11

    Parameter Efficient Fine Tuning, LMM Model Compression

12. WEEK 12

    Agentic AI and Agentic AI Frameworks

13. WEEK 13

    Model Context Protocol (MCP)

14. WEEK 14

    Multi-Modal LLMs

15. WEEK 15

    Review and Presentations

ASSESSMENT

• Rate of Midterm Exam to Success50%
• Rate of Final Exam to Success50%

READING

• - Jurafsky, D., & Martin, J. H. (2025). Speech and Language Processing (3rd ed., draft). Stanford University. Available at: http://web.stanford.edu/~jurafsky/slp3/ - Hands-On Large Language Models by Jay Alammar, Maarten Grootendorst (2024) ( https://github.com/HandsOnLLM/Hands-On-Large-Language-Models ) - LLM Engineer's Handbook by Paul Iusztin, Maxime Labonne (2024) ( https://github.com/PacktPublishing/LLM-Engineers-Handbook ) - Build a Large Language Model (From Scratch) ( https://github.com/rasbt/LLMs-from-scratch ) - Hands-On Machine Learning with Scikit-Learn and TensorFlow (https://github.com/ageron/handson-ml3) - Deep Learning for Coders with fastai and PyTorch (https://github.com/fastai/fastbook) - HuggingFace NLP Course (html) - Practical Natural Language Processing (2020) (https://github.com/practical-nlp) - Natural Language Processing with Transformers, Revised Edition (2022) (https://github.com/nlp-with-transformers )

TEACHING STAFF

• Prof.Dr. Selim AKYOKUŞCOORDINATOR
• Prof.Dr. Selim AKYOKUŞ