Agentic AI Fundamentals

Instructor-led training Agentic AI Fundamentals

Over deze training

In deze intensieve 4-daagse hands-on training leer je de basis van Agentic AI: Van LLM internals, Prompt Engineering en Retrieval Augmented Generation, Skills en MCP tot en met het bouwen van Secure Custom Agents.

Duur: 4 dagen Vorm: Instructor-led, hands-on labs Taal: Nederlands of Engels Niveau: Beginner

Doelgroep

Deze training is geschikt voor Software Engineers, AI Developers, Software Architects en CTO’s.

Ervaring met Software Development is een vereiste.

Modules

Introduction to Agentic AI

  • Artificial Intelligence
  • From rule-based systems to LLMs
  • What makes an AI system “Agentic”
  • Agents vs. Chatbots vs. Copilots
  • Major Models & Providers
  • Limitations of Agentic AI
  • Agentic AI workflow examples

LLM essentials

  • Introduction to LLMs
  • Transformer architecture
  • Tokens and Embeddings
  • Pre-training vs. fine-tuning
  • Parameter size and capabilities
  • Context windows
  • Open-weight models
  • Running models locally

Prompt Engineering

  • Anatomy of an effective prompt
  • Zero-shot vs. few-shot prompting
  • Chain-of-thought & reasoning prompts
  • System prompts vs. User prompts
  • Structured output & formatting
  • Common prompting anti-patterns
  • Iterating & Testing prompts
  • Handling ambiguity and edge cases

Agents

  • The anatomy of an AI agent?
  • The agent loop: Plan, Act, Observe
  • Tools and Function calling
  • Memory: Short-term vs. Long-term
  • Planning & Task decomposition
  • Reasoning strategies
  • Agent frameworks overview
  • Autonomy levels & Human-in-the-loop
  • Common agent failure modes

Skills

  • Introduction to Agent Skills
  • The Anatomy of a Skill
  • Progressive disclosure & Context efficiency
  • Skill discovery and selection
  • Building custom skills
  • Composing and Chaining multiple skills
  • Testing, versioning & maintaining skills
  • Security risks of skills

Model Context Protocol

  • Introduction to MCP
  • Purpose and benefits of MCP
  • Clients and Servers in MCP
  • Structure of MCP messages
  • MCP endpoints
  • Adding context with MCP
  • Creating a custom MCP Server
  • MCP configuration
  • Integrate a MCP server in an IDE
  • MCP Security

Retrieval Augmented Generation

  • Introduction to RAG
  • RAG Grounding & Freshness
  • Chunking & Document preprocessing
  • Creating Embeddings
  • Vector databases
  • Indexing & Similarity metrics
  • Building a RAG pipeline
  • Hybrid retrieval
  • Evaluating RAG quality
  • RAG vs. Fine-tuning

Building custom agents

  • The anatomy of a custom agent
  • Custom instructions & Instruction files
  • Custom chat modes
  • Reusable prompt files
  • Connecting Tools & MCP servers
  • Creating a custom agent step-by-step in VS Code
  • Testing & Verifying a custom agent
  • Sharing custom agents

Multi-Agent Orchestration

  • Introduction to Multi-Agent Orchestration
  • Orchestration patterns
  • Agent-2-Agent communication & Handoffs
  • Managing shared state
  • Challenges with Multi-Agent Orchestration
  • Introduction to the Microsoft Agent Framework
  • Multi-Agent Orchestration Microsoft Agent Framework

Evaluation & Testing of Agents

  • Evaluating agents
  • Defining success criteria & evaluation metrics
  • Building evaluation datasets
  • Unit testing agent components
  • End-to-End scenario testing
  • Using LLM as a judge
  • Human in the loop review & feedback
  • Continuous evaluation in CI/CD
  • Benchmarking & comparing agent versions

Harnesses & Guardrails

  • Introduction to Agent harnesses
  • Sandboxing & execution isolation
  • Tool access control
  • Human-in-the-loop approval gates
  • Input & output Guardrails
  • Rate limiting, cost & resource controls
  • Guardrail frameworks & tooling
  • Designing safe defaults for autonomous agents

Security essentials

  • Threat modeling for agentic AI systems
  • Prompt Injection & Jailbreaking
  • Data privacy & PII handling
  • Secrets & credential management
  • Supply chain security
  • Secure handling of agent-generated code
  • Adversarial testing & Red Teaming
  • Monitoring, logging & incident response
  • Compliance considerations (GDPR, AI Act, industry standards)

Governance & Best Practices

  • AI governance principles
  • Responsible AI & ethical considerations
  • Risk assessment & Classification of agentic systems
  • Accountability for AI systems
  • Documentation & Auditability
  • Change management & Versioning
  • Best practices for production rollout

Appendix: Openspec, Spec Driven Development

Voorkennis

Basiskennis van programmeren is een vereiste.

Interesse?

Neem contact op voor meer informatie, beschikbaarheid of een offerte op maat.