| 1 | Tokenization and Text Representations Inside a Model | The prompt becomes token IDs, embeddings and contextual states. |
| 2 | Autoregressive Generation and Decoding | The model picks one next token at a time: temperature, top-k, top-p, stopping and structured output. |
| 3 | Attention Mechanism | How each token looks up the others with Q, K and V, worked through in numbers. |
| 4 | Transformers: Original Architecture and Modern Blocks | The block a model stacks many times, from the 2017 design to modern components. |
| 5 | Attention Variants and KV-Cache Compression | Why memory grows with context, and how MQA, GQA, MLA and cache compression shrink it. |
| 6 | Mixture of Experts: Sparse Computation, Routing, and Deployment Cost | A router sends each token to a few experts: active parameters, total weights, memory and speed. |
| 7 | Multimodal Models: Alignment, Fusion, and Visual Questions | How images join the input: patches, modality connectors and visual questions. |
| 8 | Encoder, Decoder, and Encoder–Decoder: How Information Flows | Which models read and which write: classifiers, generators and sequence converters. |
| 9 | Pretraining, Supervised Fine-Tuning, and Preference Optimization | How pretraining, supervised fine-tuning and preference optimization shape what the model does. |
| 10 | Transfer Learning, LoRA, and Distillation | Adapt a model to your task with a frozen encoder, full fine-tuning, LoRA or a smaller student. |