AI / LLM
LLM Bootcamp
Transformers, fine-tuning, RAG, and LLMOps.
π§ LLM Boot Camp Curriculum
π Module 1: Foundations of Large Language Models
1. What is a Language Model?
- Evolution from N-grams to Transformers
- Why βlargeβ matters in modern AI
2. Transformers 101
- Self-attention mechanism
- Positional encoding & architecture
- Mini Transformer demo in PyTorch
π Module 2: Pre-trained LLMs β Under the Hood
3. Tokenization Deep Dive
- BPE, WordPiece, SentencePiece
- Tokenizer visualization & decoding
4. Exploring Popular LLMs
- GPT, LLaMA, Falcon, Mistral
- Model card reading and comparison
5. Embeddings & Representations
- Sentence and word embeddings
- Using OpenAI or Hugging Face APIs
π Module 3: Prompting & Fine-Tuning
6. Prompt Engineering Basics
- Zero, one, few-shot prompting
- Prompt hacks & real-world examples
7. Fine-Tuning vs LoRA
- Why and when to fine-tune
- Fine-tune with PEFT / LoRA on Hugging Face
8. RLHF (Reinforcement Learning from Human Feedback)
- Overview of alignment techniques
- Human feedback simulation demo
π Module 4: Building with LLMs
9. LangChain / LlamaIndex 101
- Chains, memory, and agents
- Building Q&A apps with RAG
10. LLM APIs in Production
- OpenAI, Cohere, HuggingFace Inference
- Handling rate limits, logging, observability
11. Chatbot + RAG Project
- Ingest PDF/text
- Add memory & tool use
π Module 5: Advanced Topics & Deployment
12. LLMOps: Serving & Scaling
- FastAPI, vLLM, Hugging Face Spaces, Modal
- Deploying local models with Ollama
13. Security, Bias & Ethics
- Prompt injection attacks
- Jailbreaks, red-teaming & safe responses
14. Build a Mini Language Model
- Train from scratch on a toy dataset
- Tokenization β model β sampling loop
π Capstone Projects (Pick One)
- β Chatbot for documents (PDF, CSV)
- β Auto-tagging for customer support
- β Custom-trained LoRA chatbot
- β RAG-based search engine
π Bonus Chapter: Becoming an LLM Engineer
- Must-have skills & GitHub project ideas
- Resume and interview tips
- Building your AI portfolio
Built for developers, students, and tech teams ready to master LLMs in record time.