Course Outline
Day 1
Anatomy of a Modern AI Agent
Moving beyond chatbots: understanding agents as autonomous systems for reasoning and action.
Exploring reactive, proactive, hybrid, and goal-directed agent paradigms.
Core components: perception, planning, memory, tool use, and action.
Evaluating design tradeoffs between single-agent and multi-agent systems.
Agent Frameworks and the Modern Stack
Overview of LangChain, LlamaIndex, AutoGen, and CrewAI, along with their respective tradeoffs.
Comparison with classical frameworks like JADE and SPADE.
Criteria for selecting a framework based on production requirements.
Understanding tool calling, function calling, and structured outputs.
Hands-on: Scaffolding a single Python agent equipped with tool calls.
Multi-Agent System Architectures
Design patterns for Multi-Agent Systems (MAS): centralized, decentralized, hybrid, and layered approaches.
FIPA ACL, message-passing mechanisms, and their modern equivalents.
Coordination patterns including planning, negotiation, and synchronization.
Understanding emergent behavior and self-organization within agent populations.
Decision-Making and Learning in Agents
Applying game theory to cooperative and competitive agent interactions.
Implementing reinforcement learning within multi-agent environments.
Facilitating transfer learning and knowledge sharing across agents.
Establishing conflict resolution mechanisms and trust among coordinating agents.
Day 2
Multi-Modal Foundations for Agents
The role of multi-modal AI in creating unified workflows across text, image, speech, and video.
Examining leading multi-modal models: GPT-4 Vision, Gemini, Claude, and Whisper.
Techniques for fusing different modalities within an agent's reasoning loop.
Evaluating latency, cost, and accuracy tradeoffs in multi-modal pipelines.
Building the Perception Layer
Image processing capabilities for agents: classification, captioning, and object detection.
Implementing speech recognition using Whisper ASR and streaming transcription.
Utilizing text-to-speech synthesis for natural voice interactions.
Integrating perception outputs with LLM-driven reasoning and tool selection.
Hands-On - Building a Multi-Modal Agent in Python
Defining the agent's task, context window, and tool inventory.
End-to-end integration of GPT-4 Vision and Whisper APIs.
Implementing memory management, state handling, and conversation flows.
Safely implementing tool calls that produce real-world side effects.
Hands-On - Orchestrating a Multi-Agent System
Composing specialized agents using AutoGen or CrewAI.
Defining roles, responsibilities, and inter-agent communication protocols.
Managing resource allocation and coordination in a simulated environment.
Logging agent reasoning, tool calls, and decisions for inspection and audit trails.
Day 3
Threat Surface of Production AI Agents
Understanding why agentic AI is uniquely vulnerable compared to traditional software.
Mapping the attack surface across data, model, prompt, tool, output, and interface layers.
Conducting threat modeling for agent-based systems with autonomous tool use.
Comparing AI cybersecurity practices with traditional cybersecurity methods.
Adversarial Attacks Hands-On
Exploring adversarial examples and perturbation methods: FGSM, PGD, and DeepFool.
Differentiating between white-box and black-box attack scenarios.
Analyzing model inversion and membership inference attacks.
Addressing data poisoning and backdoor injection during the training phase.
Tackling prompt injection, jailbreaking, and tool misuse in LLM-based agents.
Defensive Techniques and Model Hardening
Implementing adversarial training and data augmentation strategies.
Utilizing defensive distillation and other robustness techniques.
Applying input preprocessing, gradient masking, and regularization methods.
Incorporating differential privacy, noise injection, and privacy budget management.
Leveraging federated learning and secure aggregation for distributed training.
Hands-On with the Adversarial Robustness Toolbox
Simulating attacks against the multi-modal agent constructed on Day 2.
Measuring 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
NIST AI Risk Management Framework: govern, map, measure, manage.
ISO/IEC 42001 and emerging AI-specific standards.
Mapping AI risks to existing enterprise GRC frameworks.
Requirements for AI accountability, auditability, and documentation.
Regulatory Compliance for Agentic Systems
EU AI Act: risk tiers, prohibited uses, and obligations for high-risk systems.
Implications of GDPR and CCPA for agent data pipelines.
U.S. Executive Order on Safe, Secure, and Trustworthy AI.
Sector-specific guidance for finance, healthcare, and public services.
Managing third-party risk and supplier AI tool usage.
Ethics, Bias, and Explainability
Detecting and mitigating bias across agent perception and reasoning processes.
Recognizing explainability and transparency as critical security properties.
Ensuring fairness, preventing downstream harm, and promoting responsible deployment.
Designing inclusive and auditable agent behaviors.
Production Deployment, Monitoring, and Incident Response
Secure deployment patterns for single and multi-agent systems.
Continuous monitoring for drift, anomalies, and abuse.
Maintaining logs, audit trails, and forensic readiness for agent actions.
AI security incident response playbooks and recovery strategies.
Case studies of real-world AI breaches and lessons learned.
Capstone and Synthesis
Reviewing the multi-modal multi-agent system built throughout the course.
Conducting an end-to-end pipeline review: design, build, secure, govern, deploy.
Self-assessing the system against NIST AI RMF functions.
Exploring emerging trends in agentic AI and AI security.
Summary and Next Steps
Requirements
Target Audience
This course is designed for AI engineers and architects developing agentic systems for production environments. It also caters to cybersecurity, risk management, and compliance professionals tasked with ensuring AI assurance in highly regulated sectors such as finance, healthcare, and consulting. Additionally, senior developers and solution leads who are integrating multi-modal and multi-agent capabilities into enterprise platforms will find this training essential.
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
