Module “Design and Development of AI Agents”

Topic title: Introduction to AI Agents and the Agentic Approach to Software Development.

Topic content: Differences among a chatbot, an AI assistant, and an AI agent. The role of LLMs, tools, memory, planners, context, and external data sources. Fundamental principles of the agentic approach to software development (agentic software engineering). Examples of AI agent applications in education, business, research, customer support, and process automation. 

Topic title: AI Agent Design

Topic content: Defining the goal, role, and tasks of an AI agent. Designing inputs, outputs, processing steps, and decision logic. Selecting tools, data sources, and modes of user interaction. Core elements of AI agent architecture: LLM component, system prompt, memory, tools, validation, and structured output. 

Topic title: AI Agent Development Using the LangChain and LangGraph Frameworks.

Topic content: Introduction to the LangChain and LangGraph frameworks for AI agent development. Connecting LLMs with tools, data, and processing steps. Creating a simple agentic workflow and defining states, nodes, and transitions. Practical development of an AI agent. 

Topic title: Integration, Evaluation, and Responsible Use of AI Agents.

Topic content: Integration of an AI agent with existing applications, services, and external data sources. Monitoring agent operation, testing results, and evaluating response quality. Basic output validation, identification of errors, and recognition of AI agent limitations. Principles of responsible use, security, transparency, and ethical use of AI agents. 

Learning outcomes:

Upon completion of this module, participants will be able to:

• Explain the fundamental concepts of AI agents, including the differences among a chatbot, an AI assistant, and an AI agent, as well as the role of LLMs, tools, memory, context, and external data sources. 

• Design the architecture of an AI agent by defining its goal, role, inputs, outputs, processing steps, decision logic, and modes of user interaction. 

• Develop a simple AI agent using contemporary tools and frameworks, such as LangChain, LangGraph, and similar tools, connecting LLMs with data, tools, and structured workflows. 

• Integrate, test, and evaluate an AI agent within a specific digital solution, applying basic output validation, monitoring response quality, and recognizing the limitations of agentic systems. 

• Apply principles of responsible and secure use of AI agents, including transparency, data protection, risk control, ethical considerations, and critical assessment of results generated by the AI agent. 

Literature:

Electronic materials available at https://moodle.elab.rs