Technologies

Agentic AI — Building Autonomous AI Agents

This training explores the rapidly evolving field of agentic AI — systems where large language models act as autonomous agents capable of planning, tool use, and multi-step reasoning to accomplish complex goals. Participants will learn the architectural patterns behind agent systems, how to build agents using leading frameworks such as LangChain and AutoGen, and how to design reliable, observable, and safe agentic workflows for real-world applications. The course combines conceptual foundations with hands-on coding labs.

Why choose this training?

Agentic AI is moving from research curiosity to production reality, and the gap between developers who understand agent architectures and those who do not is growing fast. This training gives software engineers and AI practitioners the conceptual models and practical coding skills to build, evaluate, and deploy agentic systems — from single-agent ReAct loops to coordinated multi-agent workflows with real-world tool integrations.

The course is deliberately grounded in production concerns, not just prototyping, so participants understand not only how to build agents but how to make them reliable, observable, and safe in organizational environments.

What makes our approach unique?

EITT brings together over 500 specialists and draws on insights from more than 2,500 trainings delivered in AI, software engineering, and cloud architecture. Our agentic AI curriculum is continuously updated to reflect the fast-moving landscape of frameworks and models, so participants always train on current tools — LangGraph, AutoGen, CrewAI — rather than last year’s approaches. This commitment to currency is what sets EITT apart in an area where the technology evolves month by month.

Benefits

Understand the architectural principles that distinguish agentic AI from standard LLM applications
Apply ReAct and other agent reasoning patterns in practical implementations
Build functional AI agents using LangChain, LangGraph, AutoGen, and CrewAI
Design effective tool integrations that allow agents to interact with external systems
Implement memory and state management for long-running agent workflows
Construct multi-agent systems with defined roles and coordination mechanisms
Apply safety, reliability, and observability best practices to production agent systems
Evaluate when agentic AI is appropriate and how to scope agent-based solutions

Who is this training for?

Software engineers and backend developers working with AI systems

Machine learning engineers expanding into agent architectures

AI/ML architects designing production AI solutions

Technical leads evaluating agentic AI for enterprise use cases

Product engineers building AI-powered applications

Data scientists looking to move beyond single-model pipelines

Prerequisites

Proficiency in Python programming
Basic familiarity with large language models and prompt engineering
Experience with REST APIs and JSON (required for tool integration labs)
Basic understanding of vector databases is helpful but not required

Training program

What makes an AI system agentic — perception, planning, action
From prompt engineering to autonomous agents — the evolution
Large language models as reasoning engines — strengths and limits
Agent taxonomies — reactive, deliberative, and hybrid agents
Key challenges — hallucination, reliability, safety, and control
Overview of the agentic AI ecosystem and tooling landscape

ReAct (Reason + Act) — the foundational agent loop
Tool use — function calling, APIs, and external integrations
Memory systems — short-term context, long-term memory, vector stores
Planning architectures — chain-of-thought, tree-of-thought, MCTS
Multi-agent systems — roles, communication, and orchestration
Agent evaluation — benchmarks and quality metrics

LangChain fundamentals — chains, agents, and tools
LangGraph — building stateful, graph-based agent workflows
Defining tools and connecting agents to external systems
Managing agent state and conversation memory
Building a research agent with web search and document retrieval
Debugging and tracing agent execution with LangSmith

AutoGen framework — conversable agents and group chat
Defining agent roles, personas, and collaboration patterns
CrewAI — crew orchestration and task delegation
Human-in-the-loop patterns — when and how to involve humans
Handling agent disagreements and conflict resolution strategies
Practical multi-agent workflow — automated research and report writing

Guardrails — input/output validation and content filtering
Handling agent failures — retries, fallbacks, and circuit breakers
Prompt injection and adversarial input defense
Observability — tracing, logging, and monitoring agent runs
Cost management — token budgets and efficiency strategies
Testing agentic systems — unit tests, integration tests, and red-teaming

Delivery Methods

Online

Convenience of participating from anywhere
Interactive live sessions with trainer
Materials available for 30 days
No travel costs

On-site

Direct contact with trainer and group
Intensive hands-on workshops
Networking with other participants
Full focus on learning

Training Topics

Frequently asked questions

What level of AI or ML knowledge is needed to attend this training?

Participants should have basic familiarity with large language models and prompt engineering — for example, having used the OpenAI API or similar. Deep ML or data science knowledge is not required. The training focuses on building agentic systems using high-level frameworks rather than training models from scratch.

How are the 2 days structured?

Day 1 covers agentic AI foundations, architecture patterns, and building single agents with LangChain and LangGraph, with hands-on labs throughout. Day 2 focuses on multi-agent systems (AutoGen, CrewAI), production readiness topics, and a capstone lab where participants build a complete multi-agent workflow. Both online and onsite formats are available.

Which AI frameworks and models are used in the labs?

The hands-on labs primarily use LangChain, LangGraph, AutoGen, and CrewAI as agent frameworks. Labs use OpenAI GPT-4o as the underlying LLM by default, though the architectures and patterns are model-agnostic and applicable to other providers such as Anthropic Claude or open-source models.

Is this training relevant for enterprise use cases or just for experimentation?

This training explicitly addresses production concerns — reliability, safety, observability, and cost management — making it directly applicable to enterprise use cases. Module 5 is dedicated to making agentic systems production-ready, covering guardrails, testing strategies, and monitoring. Participants leave with patterns applicable to real organizational deployments.

Why choose EITT for this training?

EITT is a training provider with over 500 experts and experience from over 2,500 trainings delivered in AI, cloud, and software engineering. Our agentic AI trainers are active practitioners building production agent systems, ensuring the curriculum reflects the current state of the field rather than outdated or purely academic content. EITT holds a 4.8/5 average participant rating.