Foundation Week

Introduction & Setup

Start with a smart home demo to understand what agents are, then explore the frameworks landscape, career paths, and IDE setup for Windows and Mac.

Core Design Patterns

Pattern	Description
Agent Autonomy	Agents make decisions independently based on goals
Sequential Workflow	Step-by-step chain of LLM calls
Routing	Direct input to specialized sub-agents
Parallelization	Fan-out to multiple models simultaneously

Multi-LLM Orchestration

Compare GPT-4o, Claude, Gemini, DeepSeek. Each has strengths — route tasks to the right model.

# Multi-LLM router example
async def route_task(task_type, prompt):
    if task_type == "vision":
        return await call_gpt4o(prompt)
    elif task_type == "code":
        return await call_claude(prompt)
    elif task_type == "reasoning":
        return await call_deepseek(prompt)
    else:
        return await call_gemini(prompt)

Frameworks & UI Tools

OpenAI Agents SDK — Lightweight, Python-native agent framework
CrewAI — Role-based multi-agent collaboration
LangGraph — Graph-based stateful agent orchestration
AutoGen — Multi-agent conversation framework from Microsoft
Gradio — Rapid UI prototyping for agent demos

Deployment & Wrap-up

Learn LLM function calling, tool handling, AI assistant patterns, and deploying agents to production. End with a career chatbot project.

✏️ Exercise: Set up your development environment (Python 3.11+, VS Code, Git). Build a simple multi-step LLM workflow that takes a user query, classifies it (technical, billing, general), and routes to a specialized prompt. Use at least two different LLM providers.