Skip to content

curriculum.md

Latest commit

 

History

History
98 lines (79 loc) · 3.3 KB

curriculum.md

File metadata and controls

98 lines (79 loc) · 3.3 KB

Practical LLM Course — 9-Session Curriculum

Target Audience

Master students with medium programming proficiency. Each session is 40 minutes of hands-on coding using the OpenAI API (Chat Completions, Embeddings, Function Calling).

Session 1: LLM API Fundamentals

Objective: Set up the OpenAI Python client and understand the core Chat Completions API.

  • Installing the openai package
  • API key configuration
  • First chat completion request
  • Understanding message roles: system, user, assistant
  • Key parameters: temperature, max_tokens, top_p
  • Streaming responses
  • Exercise: Build a simple language translator

Session 2: Prompt Engineering in Practice

Objective: Master practical prompting techniques that improve LLM output quality.

  • Zero-shot prompting
  • Few-shot prompting (1-shot, multi-shot)
  • Chain-of-Thought (CoT) prompting
  • System prompt design patterns
  • Prompt templates with Python f-strings
  • Exercise: Build a sentiment analyzer with different prompting strategies

Session 3: Structured Outputs & Output Parsing

Objective: Extract structured data from LLMs reliably using JSON mode and schemas.

  • JSON mode (response_format: json_object)
  • Structured outputs with json_schema
  • Parsing and validating LLM outputs
  • Error handling for malformed outputs
  • Exercise: Build a resume/CV information extractor

Session 4: Text Embeddings & Semantic Search

Objective: Understand embeddings and build a semantic search engine.

  • What are text embeddings?
  • OpenAI Embeddings API (text-embedding-3-small)
  • Cosine similarity computation
  • Building a semantic search system over documents
  • Exercise: Search a FAQ knowledge base

Session 5: Retrieval-Augmented Generation (RAG)

Objective: Build a complete RAG pipeline to answer questions from custom documents.

  • RAG architecture overview
  • Document chunking strategies
  • Embedding and indexing documents
  • Retrieval + generation pipeline
  • Exercise: Q&A chatbot over a custom document set

Session 6: Function Calling & Tool Use

Objective: Enable LLMs to call external functions and use tools.

  • What is function calling?
  • Defining tool schemas
  • Handling function call responses
  • Multi-tool scenarios
  • Exercise: Build an assistant with calculator and weather tools

Session 7: Building LLM Agents

Objective: Build an autonomous agent that reasons and acts using tools in a loop.

  • What is an LLM agent?
  • The agent loop (Observe → Think → Act)
  • ReAct-style reasoning
  • Conversation memory
  • Exercise: Build a research agent that can search and calculate

Session 8: LLM Application Patterns & Best Practices

Objective: Learn production patterns for robust LLM applications.

  • Chaining multiple LLM calls
  • Input guardrails and safety
  • Token counting and cost management
  • Basic LLM output evaluation
  • Exercise: Build a moderated content pipeline

Session 9: Capstone — Full LLM Application

Objective: Combine all concepts into a complete, functional LLM application.

  • Architecture review: RAG + Tools + Agent
  • Building a knowledge-base powered assistant
  • Adding tools (search, calculator)
  • Interactive conversation loop
  • Course recap and next steps
  • Exercise: Extend the assistant with a new tool or knowledge source