Course Overview
This beginner-level course teaches how to build working large language model (LLM) applications including prompts, structured outputs, retrieval-augmented generation (RAG), tool calling, and AI agents. It covers the entire AI engineering stack from understanding how LLMs generate text to building, testing, securing, and improving LLM apps. The course includes practical projects such as building a RAG chatbot and a research agent, and emphasizes testing with eval sets and guarding against hallucinations and prompt injection.
Key Takeaways
- Understand the layers of an LLM app and how LLMs generate text token by token.
- Call LLM APIs reliably with chat history, streaming, retries, and secure API key management.
- Write system and reasoning prompts that return schema-valid JSON from text, images, and PDFs.
- Extract, classify, and summarize large document sets with validation and confidence routing.
- Build and evaluate RAG chatbots using vector databases, hybrid search, reranking, and citations.
- Create tool-calling AI agents with memory, step limits, spend caps, and human approval workflows.
- Test LLM apps using eval sets, LLM judges, agent path evaluations, and continuous integration.
- Implement observability, data redaction, prompt injection defenses, and cost optimization techniques.
- Improve prompts with user feedback loops and A/B testing.
- Gain an overview of advanced topics like multi-agent systems, fine-tuning, voice and image generation, and production deployment.
Prerequisites
- A computer with Python 3 installed and an API key from an LLM provider.
- Basic Python knowledge including variables, functions, lists, and dictionaries.
- No prior machine learning, math, or framework experience required.
Target Learners
- Beginners who want to understand and build LLM apps, RAG, and AI agents.
- Developers, students, analysts, and product professionals starting in AI engineering.
- Not suitable for those already shipping RAG or agents in production or seeking model training, fine-tuning, or MLOps knowledge.
Key terms: token, context window, temperature, API 2:26
The AI engineering stack: every layer of an LLM app mapped 7:04
Next-token prediction: why LLMs guess instead of look up 6:29
Temperature and sampling: why the same prompt changes 6:01
Tokens and context windows: estimate what fits in a call 5:48
Your first LLM API call: send a prompt, read the response 6:00
Multi-turn chat: give an LLM memory with message history 6:21
Streaming, rate limits and retries: make API calls reliable 6:47
Run open LLMs on your laptop with Ollama 8:11
Choose an LLM provider and set up API keys safely 8:04
How LLMs actually work
Price and run one API call
Key terms: system prompt, few-shot, JSON Schema 2:03
System prompts and few-shot examples that steer an LLM 5:26
Structured outputs: get JSON your code can actually trust 4:54
Practice: fix 10 broken prompts with one checklist 4:12
Reasoning prompts: make an LLM work step by step and check itself 6:46
Vision prompts: send images and PDFs to an LLM 6:16
Prompt templates and versioning: manage prompts like code 7:26
Extraction and classification at scale: process thousands of documents 7:57
Summarization that keeps the facts: long documents, short and correct 7:38
Prompts and structured outputs
A schema-valid extraction call
Key terms: embedding, chunk, vector search, RAG 2:05
Embeddings: how semantic search matches meaning, not words 6:11
Chunking documents for RAG without losing the context 4:51
Build a RAG pipeline: retrieve, augment, then generate 5:08
RAG debugging: wrong chunk, missing chunk, or bad answer 4:43
Vector databases: store and search embeddings at scale 6:51
Hybrid search and reranking: put the right chunk first 6:34
Project: build a RAG chatbot over your PDFs 7:49
Evaluate RAG: measure retrieval and answers separately 7:54
RAG
Index, ask, label
Key terms: tool, tool_use, agent loop, guardrail 2:17
Tool calling: let an LLM use functions your code runs 4:56
The agent loop: how AI agents think, act and observe 5:03
Prompt chaining vs AI agents: pick the simplest design 4:53
Agent guardrails: step limits, spend caps, human approval 4:48
Workflow patterns: routing, parallel calls and evaluator loops 6:24
Agent memory: what an AI agent should remember between sessions 6:42
Project: build a research agent with web search 7:31
Agent frameworks vs plain code: LangChain, LlamaIndex and SDKs 7:51
Tool calling and AI agents
Guard an agent loop
Key terms: eval, pass rate, judge, prompt injection 2:08
Hallucinations: spot invented facts before users do 4:40
LLM evals: build a test set that catches bad answers 5:18
LLM-as-a-judge: grade answers, then check the grader 5:07
Prompt injection: how AI apps get hijacked and defended 5:10
LLM observability: log and trace every call your app makes 6:21
PII and content guardrails: filter what goes in and out 6:52
Cut LLM cost and latency with caching, batching and smaller models 6:34
Run evals in CI: test every prompt change automatically 7:47
Evaluate AI agents: grade the path, not just the answer 8:02
User feedback and A/B tests: improve prompts with real users 7:36
Capstone: ship a document Q&A assistant with an agent 7:02
Testing, securing and improving LLM apps
Eval cases for your capstone
Key terms: MCP server, context rot, coding agent 2:10
MCP: connect any AI app to tools with one protocol 4:45
Context engineering: what belongs in an agent's context 5:05
Reasoning, multimodal or open models: choose the right LLM 5:24
Vibe coding with AI coding agents: where it breaks 4:51
Multi-agent systems: when many AI agents beat one 6:38
Fine-tuning vs RAG vs prompting: pick the right fix 6:22
Ship an LLM app to production: keys, limits and fallbacks 6:15
Voice AI apps: speech-to-text, LLM and text-to-speech 7:52
Image generation in apps: prompts, costs and safety 7:59
Computer-use and browser agents: what they can and can't do 8:02
Your AI engineering portfolio: three projects that prove your skills 8:16
MCP, context engineering, coding agents
Map an AI app to MCP
Access Files No pi required.
Free Course files 1 file package
- Full Pack
Discussions are closed
Comments are currently disabled for this course.
