GenAI & LLM Engineering — 2-Week Crash Training¶
Ten sessions. Build production-ready LLM systems. Walk into your next interview with working projects and real evaluation numbers.
What You'll Be Able to Do¶
After completing both weeks, you will be able to:
- Explain how attention, tokenization, and context windows work — and why they constrain system design
- Build and evaluate RAG pipelines: chunking, retrieval, reranking, and RAGAS metrics
- Write production FastAPI endpoints with async patterns, Server-Sent Events streaming, and caching
- Design and implement LangGraph agents with conditional routing and human-in-the-loop checkpoints
- Fine-tune a model with QLoRA, merge the adapter weights, and serve it on CPU
- Design LLM systems under constraints: latency, cost, accuracy, and privacy tradeoffs
Course Structure¶
| Day | Part 1 | Part 2 |
|---|---|---|
| 01 | LangChain Fundamentals | Advanced RAG |
| 02 | Fine-Tuning | Function Calling & Tool Use |
| 03 | AI Agents | LangGraph |
| 04 | LLMOps | Deployment |
| 05 | Capstone Project | Mock Interview & Portfolio |
6 Guided Projects¶
Each project ships with a setup guide, full implementation, advanced features, evaluation scripts, deployment instructions, and interview Q&A.
| # | Project | What You Build | Core Skills |
|---|---|---|---|
| 1 | RAG Q&A Chatbot | PDF ingestion → embedding → retrieval → streaming answers with citations | ChromaDB, CrossEncoder reranking, RAGAS |
| 2 | AI Writing Assistant | 4-stage LCEL pipeline: outline → draft → refine → style check | LangChain, async streaming, SSE |
| 3 | Document Summarizer | Map-reduce summarization with faithfulness and coherence evaluation | asyncio.gather, LLM-as-judge |
| 4 | Function-Calling Extractor | Extract typed, validated data structures from unstructured text | OpenAI tools, Pydantic, schema generation |
| 5 | LangGraph Research Agent | Planner → researcher → writer → critic loop with quality gating | LangGraph StateGraph, conditional routing |
| 6 | Fine-Tuned Classifier | QLoRA fine-tune a small model, merge weights, compare vs prompt-only | LoRA, PEFT, SFTTrainer, CPU serving |
Quick Start¶
git clone https://github.com/KirkYagami/2-Week-GenAI-LLM-Engineering-Crash-Training.git
cd 2-Week-GenAI-LLM-Engineering-Crash-Training
pip install -r requirements.txt
cp .env.example .env # add your OPENAI_API_KEY and ANTHROPIC_API_KEY
mkdocs serve # open http://127.0.0.1:8000
What API keys do you need?
An OpenAI API key covers most of the course. An Anthropic key is used in Week 01 Day 02. LangSmith is optional but recommended for Week 02 tracing exercises — it's free to sign up.
Reference Materials¶
Cheat Sheets — one-page code references for every major tool
OpenAI API · Anthropic API · LangChain LCEL · LangGraph · ChromaDB · Prompt Engineering · RAG · Fine-Tuning · AI Agents · LLMOps
Interview Preparation
Overview & Study Plan · Mock Interview Simulator
Resources
Essential Papers · Courses · YouTube Channels · Practice Platforms
Other
Glossary · Progress Tracker · Companion Notebooks
The point of this course
LLM engineering is a systems discipline. The goal is not to know the most papers — it is to build systems that work reliably, evaluate them honestly, and explain your decisions clearly. Every note, project, and exercise here is in service of that.