Introduction to Multi-Agent Systems

AI Agents with Hugging Face smolagents

Adel Nehme

VP of AI Curriculum, DataCamp

Scenario: Career Advisor Agent

The request involves many tasks:

Updating resumes
Job searching
Interview prep
Salary research

Each task needs different tools and workflows!

Why Use Multi-Agent Systems?

Use a team of specialized agents
Each agent stays focused on one task or domain
Prevents overload and confusion

A Specialized Resume Agent

resume_agent = CodeAgent(
    tools=[WebSearchTool(), skill_translator, layout_generator],
    model=InferenceClientModel(),
    instructions="You are an expert in everything related to resumes.",
    name="resume_agent",
    description="Expert in resume writing and skill translation for career transitions"
)

A Company Research Specialist

company_agent = CodeAgent(
    tools=[WebSearchTool(), background_compatibility_checker],
    model=InferenceClientModel(),
    instructions="You are an expert in everything related to company research",
    name="company_agent",
    description="Expert in researching companies, culture, and hiring practices for job seekers"
)

Create additional specialized agents using the same pattern.

The Manager Agent

career_manager = CodeAgent(
    tools=[],
    model=InferenceClientModel(model_id="deepseek-ai/DeepSeek-R1"), # Reasoning model
    instructions="You are an advisory agent to help professionals build stellar careers",
    managed_agents=[resume_agent, company_agent, interview_agent, salary_agent]
)

Delegates tasks to the appropriate specialists based on their descriptions.
Coordinates four specialist agents using the managed_agents parameter.
Benefit from models with strong reasoning and coordination capabilities.

Multi-Agent Orchestration

result = career_manager.run("I want to switch from marketing to data science. 
Help me update my resume, find companies hiring, prepare for interviews, and understand salaries.")

Let's practice!

AI Agents with Hugging Face smolagents