Add Chapters 11–15: complete the Practitioner Track

README.md

@@ -56,11 +56,11 @@ Apply what you've learned to real-world machine learning and AI problems.
 | 8 | [Unsupervised Learning: Clustering & Dimensionality Reduction](./chapters/chapter-08-unsupervised-learning/) | 8h | ✅ Available |
 | 9 | [Deep Learning Fundamentals](./chapters/chapter-09-deep-learning-fundamentals/) | 12h | ✅ Available |
 | 10 | [Natural Language Processing Basics](./chapters/chapter-10-natural-language-processing-basics/) | 8–10h | ✅ Available |
-| 11 | Large Language Models & Transformers | 10h | 🔄 Coming Soon |
-| 12 | Prompt Engineering & In-Context Learning | 6h | 🔄 Coming Soon |
-| 13 | Retrieval-Augmented Generation (RAG) | 8h | 🔄 Coming Soon |
-| 14 | Fine-tuning & Adaptation Techniques | 8h | 🔄 Coming Soon |
-| 15 | MLOps & Model Deployment | 8h | 🔄 Coming Soon |
+| 11 | [Large Language Models & Transformers](./chapters/chapter-11-large-language-models-and-transformers/) | 10h | ✅ Available |
+| 12 | [Prompt Engineering & In-Context Learning](./chapters/chapter-12-prompt-engineering-and-in-context-learning/) | 6h | ✅ Available |
+| 13 | [Retrieval-Augmented Generation (RAG)](./chapters/chapter-13-retrieval-augmented-generation/) | 8h | ✅ Available |
+| 14 | [Fine-tuning & Adaptation Techniques](./chapters/chapter-14-fine-tuning-and-adaptation/) | 8h | ✅ Available |
+| 15 | [MLOps & Model Deployment](./chapters/chapter-15-mlops-and-model-deployment/) | 8h | ✅ Available |
 
 ### Advanced & Specialization Track (Master Complex Topics)
 Dive deep into cutting-edge techniques and specialized domains.
@@ -268,12 +268,12 @@ pie title Curriculum Breakdown
     "Community Requested" : 999
 ```
 
-- **Chapters Available Now**: 9 (76 hours of content)
+- **Chapters Available Now**: 15 (116 hours of content) — Foundation + Practitioner tracks complete
 - **Total Planned Chapters**: 25+
-- **Jupyter Notebooks**: 21 interactive notebooks
-- **SVG Diagrams**: 21 professional diagrams
-- **Exercises**: 37 problems with solutions
-- **Datasets**: 5 practice datasets
+- **Jupyter Notebooks**: 45 interactive notebooks
+- **SVG/Mermaid Diagrams**: 36 professional diagrams
+- **Exercises**: 60+ problems with solutions
+- **Datasets**: 30+ practice datasets
 - **Community-Requested Chapters**: Growing daily
 
 ---
@@ -370,5 +370,5 @@ Every share helps more people learn AI. Thank you! 🙏
 
 **Created by Luigi Pascal Rondanini | Generated by Berta AI**
 
-*Last Updated: March 2026*  
+*Last Updated: May 2026*  
 *All chapters maintained and continuously improved based on community feedback.*

ROADMAP.md

@@ -8,11 +8,11 @@ Our vision for the future of AI education. This is a living document—prioritie
 
 **Master Repository**: ✅ Live  
 **Foundation Track**: ✅ Complete (5 chapters available)  
-**Practitioner Track**: 🔄 In progress (4 of 10 chapters available)  
+**Practitioner Track**: ✅ Complete (10 of 10 chapters available)  
 **Advanced Track**: 📋 Planned (10 chapters)  
 **Community Requests**: 🚀 Starting (unlimited)  
 **Total Planned**: 25+ chapters, 500+ hours of content  
-**Currently Available**: 9 chapters, 76 hours of content, 27 SVG diagrams
+**Currently Available**: 15 chapters, 116 hours of content, 36 diagrams
 
 ---
 
@@ -21,7 +21,7 @@ Our vision for the future of AI education. This is a living document—prioritie
 ### Objectives
 - ✅ Establish master repository (DONE)
 - ✅ Complete Foundation Track (DONE)
-- ✅ Begin Practitioner Track (Ch 6-9 available)
+- ✅ Complete Practitioner Track (Ch 6-15 available)
 - 🔄 Establish community request process
 - 🔄 Build first 100 community chapters
 - ✅ Create core infrastructure and documentation (DONE)
@@ -37,11 +37,11 @@ Our vision for the future of AI education. This is a living document—prioritie
 - One new chapter released per week
 - New chapters unlock after reaching **10 newsletter subscribers**
 - ✅ Foundation Track complete (Chapters 1-5)
-- ✅ Practitioner Track started (Chapters 6-9)
+- ✅ Practitioner Track complete (Chapters 6-15)
 
 ### Metrics to Track
 - Newsletter subscribers (target: 10 to unlock weekly releases)
-- Chapters completed: 9 / 25
+- Chapters completed: 15 / 25
 - Community requests received
 - Stars on master repo
 
@@ -50,7 +50,7 @@ Our vision for the future of AI education. This is a living document—prioritie
 ## Phase 2: Practitioner Track & Community Scale
 
 ### Objectives
-- 🔄 Complete Practitioner Track (10 chapters, releasing one per week)
+- ✅ Complete Practitioner Track (10 of 10 chapters released)
 - 🔄 Scale community chapters to 50+
 - 🔄 Establish quality standards and review process
 - 🔄 Begin analytics and learner tracking
@@ -61,12 +61,12 @@ Our vision for the future of AI education. This is a living document—prioritie
 - [x] Chapter 7: Supervised Learning (Regression & Classification)
 - [x] Chapter 8: Unsupervised Learning
 - [x] Chapter 9: Deep Learning Fundamentals
-- [ ] Chapter 10: Natural Language Processing Basics
-- [ ] Chapter 11: Large Language Models & Transformers
-- [ ] Chapter 12: Prompt Engineering
-- [ ] Chapter 13: Retrieval-Augmented Generation (RAG)
-- [ ] Chapter 14: Fine-tuning & Adaptation
-- [ ] Chapter 15: MLOps & Deployment
+- [x] Chapter 10: Natural Language Processing Basics
+- [x] Chapter 11: Large Language Models & Transformers
+- [x] Chapter 12: Prompt Engineering
+- [x] Chapter 13: Retrieval-Augmented Generation (RAG)
+- [x] Chapter 14: Fine-tuning & Adaptation
+- [x] Chapter 15: MLOps & Deployment
 
 ### Infrastructure Improvements
 - [ ] GitHub Actions for automated testing

SYLLABUS.md

@@ -18,12 +18,12 @@ graph TD
     CH7["Ch 7: Supervised Learning<br/>10h | Available"]
     CH8["Ch 8: Unsupervised Learning<br/>8h | Available"]
     CH9["Ch 9: Deep Learning<br/>12h | Available"]
-    CH10["Ch 10: NLP Basics<br/>10h | Coming Soon"]
-    CH11["Ch 11: LLMs & Transformers<br/>10h | Coming Soon"]
-    CH12["Ch 12: Prompt Engineering<br/>6h | Coming Soon"]
-    CH13["Ch 13: RAG<br/>8h | Coming Soon"]
-    CH14["Ch 14: Fine-tuning<br/>8h | Coming Soon"]
-    CH15["Ch 15: MLOps<br/>8h | Coming Soon"]
+    CH10["Ch 10: NLP Basics<br/>10h | Available"]
+    CH11["Ch 11: LLMs & Transformers<br/>10h | Available"]
+    CH12["Ch 12: Prompt Engineering<br/>6h | Available"]
+    CH13["Ch 13: RAG<br/>8h | Available"]
+    CH14["Ch 14: Fine-tuning<br/>8h | Available"]
+    CH15["Ch 15: MLOps<br/>8h | Available"]
 
     CH1 --> CH2
     CH1 --> CH3
@@ -58,15 +58,15 @@ graph TD
     style CH7 fill:#4caf50,color:#fff
     style CH8 fill:#4caf50,color:#fff
     style CH9 fill:#4caf50,color:#fff
-    style CH10 fill:#f3e5f5
-    style CH11 fill:#f3e5f5
-    style CH12 fill:#f3e5f5
-    style CH13 fill:#f3e5f5
-    style CH14 fill:#f3e5f5
-    style CH15 fill:#f3e5f5
+    style CH10 fill:#4caf50,color:#fff
+    style CH11 fill:#4caf50,color:#fff
+    style CH12 fill:#4caf50,color:#fff
+    style CH13 fill:#4caf50,color:#fff
+    style CH14 fill:#4caf50,color:#fff
+    style CH15 fill:#4caf50,color:#fff
 ```
 
-**Legend**: Green = Available | Purple = Practitioner (Coming Soon) | Chapters 1-9 fully available with SVG diagrams
+**Legend**: Green = Available | Practitioner Track (Chapters 6–15) is now complete; Advanced Track (Chapters 16+) is planned
 
 ---
 
@@ -83,12 +83,12 @@ graph TD
 | 7 | [Supervised Learning](./chapters/chapter-07-supervised-learning/) | Practitioner | 10h | Available | 3 notebooks, scripts, 5 exercises, 3 SVGs |
 | 8 | [Unsupervised Learning](./chapters/chapter-08-unsupervised-learning/) | Practitioner | 8h | Available | 3 notebooks, scripts, 5 exercises, 3 SVGs |
 | 9 | [Deep Learning Fundamentals](./chapters/chapter-09-deep-learning-fundamentals/) | Practitioner | 12h | Available | 3 notebooks, scripts, 5 exercises, 3 SVGs |
-| 10 | Natural Language Processing | Practitioner | 10h | Planned | - |
-| 11 | LLMs & Transformers | Practitioner | 10h | Planned | - |
-| 12 | Prompt Engineering | Practitioner | 6h | Planned | - |
-| 13 | RAG | Practitioner | 8h | Planned | - |
-| 14 | Fine-tuning & Adaptation | Practitioner | 8h | Planned | - |
-| 15 | MLOps & Deployment | Practitioner | 8h | Planned | - |
+| 10 | [Natural Language Processing](./chapters/chapter-10-natural-language-processing-basics/) | Practitioner | 10h | Available | 3 notebooks, scripts, 4 exercises, 3 diagrams |
+| 11 | [LLMs & Transformers](./chapters/chapter-11-large-language-models-and-transformers/) | Practitioner | 10h | Available | 3 notebooks, scripts, 4 exercises, 3 diagrams |
+| 12 | [Prompt Engineering](./chapters/chapter-12-prompt-engineering-and-in-context-learning/) | Practitioner | 6h | Available | 3 notebooks, scripts, 4 exercises, 3 diagrams |
+| 13 | [RAG](./chapters/chapter-13-retrieval-augmented-generation/) | Practitioner | 8h | Available | 3 notebooks, scripts, 4 exercises, 3 diagrams |
+| 14 | [Fine-tuning & Adaptation](./chapters/chapter-14-fine-tuning-and-adaptation/) | Practitioner | 8h | Available | 3 notebooks, scripts, 4 exercises, 3 diagrams |
+| 15 | [MLOps & Deployment](./chapters/chapter-15-mlops-and-model-deployment/) | Practitioner | 8h | Available | 3 notebooks, scripts, 4 exercises, 3 diagrams |
 | 16 | Multi-Agent Systems | Advanced | 10h | Planned | - |
 | 17 | Advanced RAG | Advanced | 10h | Planned | - |
 | 18 | Reinforcement Learning | Advanced | 12h | Planned | - |

chapters/chapter-11-large-language-models-and-transformers/README.md

@@ -0,0 +1,140 @@
+# Chapter 11: Large Language Models & Transformers
+
+**Track**: Practitioner | **Time**: 10 hours | **Prerequisites**: [Chapter 10: Natural Language Processing Basics](../chapter-10-natural-language-processing-basics/)
+
+---
+
+Large language models (LLMs) and the **Transformer** architecture power most of modern AI: ChatGPT, Claude, Gemini, Llama, and the embedding/RAG systems built on top of them. This chapter takes the attention and transfer-learning ideas from Chapter 10 and builds them up into a full understanding of how transformers work, how pretrained LLMs are used, and how to build real applications around them.
+
+You will implement **scaled dot-product attention**, **multi-head attention**, **positional encodings**, and a **transformer block** in pure NumPy; work with **pretrained models** (BERT, DistilBERT, GPT-style) through a graceful Hugging Face fallback; generate embeddings; explore **decoding strategies** (greedy, top-k, top-p, temperature); and study **scaling laws**, **evaluation**, and how to ship LLM-powered features.
+
+---
+
+## Learning Objectives
+
+By the end of this chapter, you will be able to:
+
+1. **Explain the Transformer architecture** — self-attention, multi-head attention, positional encoding, residuals, layer norm
+2. **Implement attention from scratch** — scaled dot-product and multi-head attention in NumPy
+3. **Distinguish encoder, decoder, and encoder–decoder models** — and pick the right family for a task
+4. **Use pretrained LLMs** — tokenize, extract embeddings, run inference with Hugging Face `transformers`
+5. **Apply LLM embeddings to downstream tasks** — similarity search and frozen-embedding classifiers
+6. **Generate text with controlled decoding** — greedy, sampling, temperature, top-k, top-p, repetition penalty
+7. **Evaluate LLMs** — perplexity, BLEU/ROUGE, win-rate, and the limits of LLM-as-judge
+8. **Design LLM-powered systems** — chunking, streaming, function calling, and the road to RAG and fine-tuning
+
+---
+
+## Prerequisites
+
+- **Chapter 10: Natural Language Processing Basics** — tokenization, embeddings, attention intuition, transfer learning
+- **Chapter 9: Deep Learning Fundamentals** — backprop, layers, optimizers, training loops
+- Comfort with NumPy, linear algebra (matmul, softmax), and basic probability
+- Optional: PyTorch for the deeper sections (the chapter runs without it)
+
+---
+
+## What You'll Build
+
+- **Mini-Transformer in NumPy** — scaled dot-product attention, multi-head attention, positional encoding, and a single encoder block you can run end-to-end
+- **Embedding service** — wrap a pretrained model (or fallback) to turn text into vectors and search by similarity
+- **Frozen-embedding classifier** — sentence embeddings + scikit-learn for a fast, strong text classifier
+- **Decoding playground** — greedy, temperature, top-k and top-p samplers operating on real logit distributions
+- **LLM application sketch** — chunking, prompt assembly, and streaming patterns that lead into Chapter 12 (Prompt Engineering) and Chapter 13 (RAG)
+
+---
+
+## Time Commitment
+
+| Section | Time |
+|---------|------|
+| Notebook 01: Transformer Architecture (attention, multi-head, positional encoding, blocks) | 3 hours |
+| Notebook 02: Pretrained LLMs (tokenizers, embeddings, classification, model selection) | 3 hours |
+| Notebook 03: Advanced LLMs (decoding, KV cache, scaling, evaluation, apps) | 2.5 hours |
+| Exercises (Problem Sets 1 & 2) | 1.5 hours |
+| **Total** | **10 hours** |
+
+---
+
+## Technology Stack
+
+- **Numerics**: `numpy`, `pandas`, `scikit-learn`
+- **Visualization**: `matplotlib`
+- **Notebooks**: `jupyter`, `ipywidgets`
+- **Optional (LLMs)**: `transformers`, `tokenizers`, `accelerate`, `datasets`, `sentencepiece`, `huggingface-hub`
+- **Optional (DL)**: `torch` for the deeper transformer/embedding sections
+
+---
+
+## Quick Start
+
+1. **Clone and enter the chapter**
+   ```bash
+   cd chapters/chapter-11-large-language-models-and-transformers
+   ```
+
+2. **Create a virtual environment and install dependencies**
+   ```bash
+   python -m venv .venv
+   .venv\Scripts\activate   # Windows
+   # source .venv/bin/activate  # macOS/Linux
+   pip install -r requirements.txt
+   # Optional, for the pretrained-LLM sections:
+   # pip install torch transformers tokenizers accelerate datasets sentencepiece huggingface-hub
+   ```
+
+3. **Run the notebooks**
+   ```bash
+   jupyter notebook notebooks/
+   ```
+   Start with `01_transformer_architecture.ipynb`, then `02_pretrained_llms.ipynb`, then `03_advanced_llms.ipynb`.
+
+---
+
+## Notebook Guide
+
+| Notebook | Focus |
+|----------|--------|
+| **01_transformer_architecture.ipynb** | From RNN limits to attention; scaled dot-product and multi-head attention in NumPy; sinusoidal positional encoding; encoder block; encoder/decoder/decoder-only families; tokenization (BPE/WordPiece) intuition |
+| **02_pretrained_llms.ipynb** | Loading pretrained models with `transformers` (with fallback); `AutoTokenizer`; extracting and visualizing embeddings; mean pooling for sentence vectors; frozen-embedding classification; choosing BERT vs RoBERTa vs DistilBERT vs GPT |
+| **03_advanced_llms.ipynb** | Decoding strategies (greedy, sampling, temperature, top-k, top-p); KV cache shapes; scaling laws; evaluation (perplexity, BLEU/ROUGE, LLM-as-judge); building LLM apps (chunking, streaming, function calling); capstone design |
+
+---
+
+## Exercise Guide
+
+- **Problem Set 1** (`exercises/problem_set_1.ipynb`) — implement scaled dot-product attention; build sinusoidal positional encoding; plot an attention heatmap; tokenize text and reason about BPE; multi-head attention shape check; compare encoder/decoder/encoder–decoder
+- **Problem Set 2** (`exercises/problem_set_2.ipynb`) — implement top-k sampling; build a tiny transformer block from scratch; compute perplexity; train an embedding-based classifier; reason about prompt vs context-window trade-offs; evaluate generations
+- **Solutions** — in `exercises/solutions/` with runnable code, explanations, and alternatives
+
+---
+
+## How to Run Locally
+
+- Use Python 3.9+ and the versions in `requirements.txt` for reproducibility.
+- The numpy-only sections (Notebook 01, large parts of 03, all Problem Set 1) require **no** transformer installs.
+- For Notebook 02 and the embedding sections, install the optional `transformers` / `torch` extras shown above.
+- Scripts in `scripts/` can be run from the chapter root; notebooks assume that root as working directory.
+
+---
+
+## Common Troubleshooting
+
+- **`transformers` not installed** — Notebooks fall back to NumPy/sklearn stubs and print a `pip install transformers` hint; install when you want the real models
+- **Hugging Face download blocked / offline** — Set `HF_HUB_OFFLINE=1` and use a locally cached model, or rely on the fallback paths in the notebooks
+- **Out-of-memory loading a large model** — Switch `MODEL_NAME` in `scripts/config.py` to `distilbert-base-uncased` or `sentence-transformers/all-MiniLM-L6-v2`
+- **CUDA/GPU** — Optional; everything runs on CPU. Set `CUDA_VISIBLE_DEVICES=""` to force CPU if a GPU is misbehaving
+- **Slow first run** — Pretrained model download can take a few minutes; subsequent runs hit the local cache
+
+---
+
+## Next Steps
+
+- **Chapter 12: Prompt Engineering** — Now that you understand how LLMs tokenize, attend, and decode, Chapter 12 turns to *steering* them: prompt patterns, few-shot, chain-of-thought, structured output, and evaluation of prompts.
+
+---
+
+**Generated by Berta AI**
+
+Part of [Berta Chapters](https://github.com/your-org/berta-chapters) — open-source AI curriculum.  
+*March 2026 — Berta Chapters*

chapters/chapter-11-large-language-models-and-transformers/assets/diagrams/multi_head_attention.mermaid

@@ -0,0 +1,12 @@
+graph LR
+    X["Input X (batch, seq, d_model)"] --> SP["Split into h heads"]
+    SP --> H1["Head 1: Attention(Q1, K1, V1)"]
+    SP --> H2["Head 2: Attention(Q2, K2, V2)"]
+    SP --> H3["..."]
+    SP --> Hh["Head h: Attention(Qh, Kh, Vh)"]
+    H1 --> C["Concat (batch, seq, d_model)"]
+    H2 --> C
+    H3 --> C
+    Hh --> C
+    C --> P["Projection Wo"]
+    P --> O["Output (batch, seq, d_model)"]

chapters/chapter-11-large-language-models-and-transformers/assets/diagrams/self_attention.mermaid

@@ -0,0 +1,11 @@
+graph LR
+    X["Input X"] --> Q["Q = X * Wq"]
+    X --> K["K = X * Wk"]
+    X --> V["V = X * Wv"]
+    Q --> S["Scores = Q * K^T / sqrt(d_k)"]
+    K --> S
+    S --> M["Optional Mask"]
+    M --> SM["Softmax"]
+    SM --> A["Attention Weights"]
+    A --> O["Output = A * V"]
+    V --> O

chapters/chapter-11-large-language-models-and-transformers/assets/diagrams/transformer_architecture.mermaid

@@ -0,0 +1,26 @@
+graph TB
+    A["Input Tokens"] --> B["Token Embedding"]
+    B --> C["+ Positional Encoding"]
+    C --> D["Encoder Block x N"]
+    D --> E["Encoder Output"]
+
+    F["Target Tokens (shifted)"] --> G["Token Embedding"]
+    G --> H["+ Positional Encoding"]
+    H --> I["Decoder Block x N"]
+    E -.->|Cross-Attention| I
+    I --> J["Linear + Softmax"]
+    J --> K["Output Probabilities"]
+
+    subgraph Encoder Block
+        D1["Multi-Head Self-Attention"] --> D2["Add & LayerNorm"]
+        D2 --> D3["Feed-Forward"]
+        D3 --> D4["Add & LayerNorm"]
+    end
+
+    subgraph Decoder Block
+        I1["Masked Multi-Head Self-Attention"] --> I2["Add & LayerNorm"]
+        I2 --> I3["Cross-Attention"]
+        I3 --> I4["Add & LayerNorm"]
+        I4 --> I5["Feed-Forward"]
+        I5 --> I6["Add & LayerNorm"]
+    end