Course Outline
Day 1
Foundations of Contemporary AI Agent Architecture
Expanding beyond static chat interfaces to define agents as autonomous systems capable of independent reasoning and execution
Examining reactive, proactive, hybrid, and goal-directed operational paradigms
Identifying core architectural components: perception, strategic planning, memory management, tool integration, and action execution
Evaluating the technical tradeoffs between single-agent and multi-agent system designs
Agent Frameworks and the Modern Technology Stack
Analyzing the capabilities and limitations of LangChain, LlamaIndex, AutoGen, and CrewAI
Conducting a comparative analysis with legacy frameworks such as JADE and SPADE
Establishing criteria for framework selection based on production-grade operational requirements
Implementing tool invocation, function calling, and structured output protocols
Practical exercise: Constructing a single Python-based agent equipped with tool-calling capabilities
Architectural Patterns for Multi-Agent Systems
Designing centralized, decentralized, hybrid, and layered multi-agent system topologies
Reviewing FIPA ACL, message-passing mechanisms, and their modern equivalents
Applying coordination patterns for planning, negotiation, and synchronization
Assessing emergent behaviors and self-organizing dynamics within agent populations
Decision-Making and Learning Mechanisms in Agents
Utilizing game theory principles for modeling cooperative and competitive agent interactions
Applying reinforcement learning techniques within multi-agent operational environments
Implementing transfer learning and cross-agent knowledge sharing protocols
Managing conflict resolution and establishing trust among coordinating agents
Day 2
Multi-Modal Foundations for Agent Operations
Integrating multi-modal AI as a unified workflow spanning text, image, speech, and video data
Evaluating leading multi-modal models: GPT-4 Vision, Gemini, Claude, and Whisper
Developing fusion techniques to combine diverse modalities within an agent’s reasoning loop
Managing tradeoffs regarding latency, operational cost, and accuracy in multi-modal pipelines
Constructing the Perception Layer
Implementing image processing capabilities for agents, including classification, captioning, and object detection
Deploying speech recognition using Whisper ASR and streaming transcription services
Utilizing text-to-speech synthesis to facilitate natural voice interaction
Linking perception layer outputs to LLM-driven reasoning processes and tool selection logic
Practical Application: Developing a Multi-Modal Agent in Python
Defining the agent’s operational tasks, context window parameters, and available tool inventory
Configuring GPT-4 Vision and Whisper APIs in an end-to-end integration
Implementing memory structures, state management, and conversation control mechanisms
Integrating tool calls that execute safe, real-world operational side effects
Practical Application: Orchestrating a Multi-Agent System
Composing specialized agents using AutoGen or CrewAI frameworks
Defining roles, responsibilities, and inter-agent communication protocols
Managing resource allocation and coordination within a simulated operational environment
Recording agent reasoning, tool invocations, and decisions to support inspection and audit requirements
Day 3
Security Threat Surface of Production AI Agents
Identifying unique vulnerabilities in agentic AI compared to traditional software applications
Mapping the attack surface across data, model, prompt, tool, output, and interface layers
Conducting threat modeling for agent-based systems with autonomous tool execution capabilities
Aligning AI cybersecurity practices with established traditional cybersecurity standards
Adversarial Attack Simulation
Generating adversarial examples and perturbation methods: FGSM, PGD, and DeepFool
Simulating white-box versus black-box attack scenarios
Testing model inversion and membership inference attack vectors
Assessing data poisoning and backdoor injection risks during the training phase
Evaluating prompt injection, jailbreaking, and tool misuse risks in LLM-based agents
Defensive Strategies and Model Hardening
Applying adversarial training and data augmentation strategies
Implementing defensive distillation and other robustness enhancement techniques
Utilizing input preprocessing, gradient masking, and regularization methods
Applying differential privacy, noise injection, and privacy budget management
Employing federated learning and secure aggregation for distributed training environments
Practical Exercise with the Adversarial Robustness Toolbox
Simulating attacks against the multi-modal agent developed in Day 2 exercises
Measuring system robustness under perturbation and quantifying performance degradation
Iteratively applying defenses and re-evaluating attack success rates
Stress-testing tool-call pathways and prompt injection vectors
Day 4
Risk Management Frameworks for AI Governance
Implementing the NIST AI Risk Management Framework: govern, map, measure, and manage functions
Aligning with ISO/IEC 42001 and emerging AI-specific standards
Integrating AI risk assessment with existing enterprise Governance, Risk, and Compliance (GRC) frameworks
Establishing requirements for AI accountability, auditability, and documentation
Regulatory Compliance for Agentic Systems
Navigating the EU AI Act: risk tiers, prohibited uses, and obligations for high-risk systems
Addressing GDPR and CCPA implications for agent data processing pipelines
Adhering to U.S. Executive Order directives on Safe, Secure, and Trustworthy AI
Applying sector-specific guidance for finance, healthcare, and public services
Managing third-party risk and supplier AI tool usage
Ethics, Bias, and Explainability in AI Systems
Detecting and mitigating bias across agent perception and reasoning processes
Establishing explainability and transparency as security-relevant system properties
Ensuring fairness, minimizing downstream harm, and promoting responsible deployment
Designing inclusive and auditable agent behavior for government and public sector contexts
Production Deployment, Monitoring, and Incident Response
Implementing secure deployment patterns for single and multi-agent systems
Establishing continuous monitoring for drift, anomalies, and abusive usage
Maintaining logging, audit trails, and forensic readiness for agent actions
Developing AI security incident response playbooks and recovery procedures
Reviewing case studies of real-world AI security breaches and deriving lessons learned
Capstone and Synthesis
Reviewing the multi-modal multi-agent system constructed throughout the course
Conducting an end-to-end pipeline review: design, build, secure, govern, and deploy
Performing a self-assessment of the system against NIST AI RMF functional requirements
Evaluating forward outlook on emerging trends in agentic AI and AI security
Summary and Next Steps
Requirements
Targeted Audience
AI engineers and architects developing agentic systems for production use in the public sector. Cybersecurity, risk, and compliance professionals responsible for AI assurance in regulated industries including finance, healthcare, and consulting. Senior developers and solution leads embedding multi-modal and multi-agent capabilities into enterprise and government platforms.
Testimonials (3)
The trainer is patient and very helpful. He knows the topic well.
CLIFFORD TABARES - Universal Leaf Philippines, Inc.
Course - Agentic AI for Business Automation: Use Cases & Integration
Good mixvof knowledge and practice
Ion Mironescu - Facultatea S.A.I.A.P.M.
Course - Agentic AI for Enterprise Applications
The mix of theory and practice and of high level and low level perspectives
