Building LLM Applications

What you will be able to do

• You will be able to describe the LLM application stack and choose models, context strategies, and generation settings that fit a use case and a budget.
• You will be able to write prompts that return reliable structured output against a schema and check them with a simple evaluation harness.
• You will be able to turn documents into embeddings and run similarity search over a vector database, choosing a chunking and indexing strategy that fits the data.
• You will be able to build a retrieval-augmented generation pipeline that grounds answers in your own sources, adds reranking, and cites where each claim came from.
• You will be able to define tool schemas and implement a function-calling loop that lets a model query APIs and act on live systems.
• You will be able to harden tool use with input validation, error recovery, and sensible selection when a model has many tools to choose from.
• You will be able to design a multi-step agent with planning and memory using an orchestration framework, rather than hand-rolling brittle control flow.
• You will be able to instrument an LLM application for cost, latency, quality, and safety, and add guardrails before it ships.

What is inside

4 modules, 8 lessons. Each module ends in a graded quiz and most carry an assignment.

1. 01

   Foundations for Building on LLMs

   Before you wire up retrieval or agents, you need a dependable way to call a model and to judge what comes back. This module builds the application stack in your head: models, tokens, context windows, and cost, then the prompting and structured-output patterns that hold up in production. You will also stand up a basic evaluation habit so every later module has a way to prove it works.

   2 lessons · 5 quiz questions

2. 02

   Embeddings, Vector Search, and RAG

   Retrieval-augmented generation is how you ground a model in your own data and cut hallucination, and it is the backbone of most real LLM products. This module walks the full pipeline: turning text into embeddings, storing and searching them in a vector database, and assembling retrieved context into grounded, cited answers. You will also learn to fix weak retrieval with better chunking and reranking instead of blaming the model.

   2 lessons · 5 quiz questions · assignment

3. 03

   Tool Use and Function Calling

   Tools let a model reach past its training data to query APIs, run code, and change live systems, which is the step that turns a chatbot into software. This module covers defining tool schemas, running the function-calling loop, and validating and handling what tools return. You will also learn to keep the loop reliable when the model picks the wrong tool or supplies bad arguments, and where an open standard like the Model Context Protocol (MCP) fits.

   2 lessons · 5 quiz questions

4. 04

   Agents and Orchestration

   Agents chain reasoning, memory, and tools into systems that pursue a goal across many steps, and orchestration frameworks give you the state and control flow to manage them. This module covers the agent design patterns you will actually reach for (ReAct, plan-and-execute, and reflection) and the framework categories that run them (LangChain, LlamaIndex, and graph-based orchestrators like LangGraph). It finishes on the production concerns that decide whether an agent is safe to ship: tracing, cost, latency, and guardrails.

   2 lessons · 5 quiz questions · assignment