AI Agents courses

• Understand what AI agents are and how agentic systems differ from traditional LLM applications, conversational chatbots, and rule-based or workflow automation systems.
• Explore the core components of agentic systems, including models, goals, instructions, tools, memory, state, environments, and mechanisms for interacting with external systems.
• Understand agent execution loops in which the system interprets a goal, reasons about the next action, executes an action, observes the result, and iterates until a completion condition is reached.
• Compare common agent architectures and identify when agentic AI is appropriate relative to simpler deterministic workflows or conventional LLM applications.

• Build complete AI agents capable of receiving goals, interpreting tasks, selecting appropriate actions, and progressing toward defined outcomes.
• Design clear system instructions, agent roles, behavioral constraints, operating boundaries, and explicit task-completion criteria.
• Implement foundational agent loops, execution flows, state management, and mechanisms for handling intermediate results across multi-step tasks.
• Develop practical agents for research, productivity, information processing, and business workflows while maintaining appropriate human control.

• Enable AI agents to interact with APIs, databases, search systems, enterprise applications, and external services so that they can perform actions beyond text generation.
• Design reliable tool schemas, function definitions, parameters, structured inputs, and structured responses that agents can interpret and use consistently.
• Implement tool selection, execution validation, error handling, retries, fallback strategies, and mechanisms for recovering from failed or incomplete tool calls.
• Build agents capable of performing real-world actions safely by combining model reasoning with controlled, validated, and permission-aware tool execution.

• Design short-term, long-term, episodic, and semantic memory mechanisms that allow agentic applications to retain and reuse information across interactions.
• Manage conversation history, user context, application state, task progress, and persistent knowledge efficiently throughout agent execution.
• Combine agent memory with retrieval systems, vector databases, and external knowledge stores when persistent or semantically searchable information is required.
• Control context growth, information relevance, privacy, token consumption, retention policies, and memory quality during long-running or repeated agent interactions.

• Design agents that transform complex user goals into structured plans, manageable subtasks, dependencies, and executable actions.
• Explore planning, reflection, routing, iteration, dynamic task decomposition, and related techniques used to organize complex agent execution.
• Build agents capable of revising or adapting their plans in response to intermediate results, failures, new information, and changing execution conditions.
• Balance model-driven autonomy with deterministic workflows, explicit rules, checkpoints, and constraints to achieve more reliable task execution.

• Combine AI agents with Retrieval-Augmented Generation systems, enterprise documents, databases, structured information, and external knowledge sources.
• Enable agents to determine dynamically when retrieval is necessary, which knowledge source should be consulted, what query should be issued, and how retrieved evidence should influence subsequent actions.
• Build iterative retrieval, investigation, and research workflows in which agents gather and refine information across multiple steps before producing an outcome.
• Improve groundedness, factual reliability, source attribution, and evidence traceability in knowledge-intensive agent-generated outputs.

See also: RAG Systems courses →

• Design systems in which specialized agents collaborate, delegate tasks, exchange information, coordinate actions, and contribute toward shared objectives.
• Explore supervisor, hierarchical, sequential, parallel, and peer-to-peer multi-agent architectures and understand how each pattern organizes responsibilities and execution.
• Define specialized agent roles, communication protocols, task-routing mechanisms, shared state, coordination rules, and conflict-resolution strategies.
• Evaluate when multi-agent systems provide meaningful advantages in specialization, parallelism, or coordination and when simpler single-agent architectures are more reliable and efficient.

• Evaluate AI agents across task completion, decision quality, tool selection, execution reliability, latency, resource consumption, and operational cost.
• Build representative test scenarios covering expected workflows, edge cases, tool failures, incomplete information, adversarial conditions, and unexpected agent behavior.
• Trace agent decisions, actions, tool calls, state transitions, intermediate outputs, failures, and complete execution paths to understand system behavior.
• Use evaluation, testing, tracing, and observability pipelines to diagnose weaknesses systematically and continuously improve agent performance and reliability.

• Understand prompt injection, excessive agency, insecure tool use, sensitive-data leakage, permission escalation, malicious inputs, and other risks specific to agentic systems.
• Implement authentication, authorization, permissions, sandboxing, input and output validation, execution constraints, guardrails, and human approval mechanisms.
• Define explicit boundaries governing which information agents may access and which decisions, transactions, or actions they may perform autonomously.
• Build auditable and governed agentic systems with traceability, monitoring, escalation mechanisms, accountability, and appropriate human oversight for enterprise environments.

See also: AI Security courses →

• Architect production-grade agentic systems that combine foundation models, tools, memory, Retrieval-Augmented Generation, workflows, APIs, databases, and application infrastructure.
• Design for scalability, reliability, fault tolerance, concurrency, latency, operational cost, long-running execution, and recovery from partial or failed agent actions.
• Implement monitoring, evaluation, versioning, tracing, deployment, incident analysis, feedback collection, and continuous-improvement practices for production agents.
• Build and deploy an end-to-end AI agent capable of reliably executing complex real-world workflows using modern AgentOps and production engineering practices.