Course Overview
This course empowers professionals to move beyond static prompt-based AI and into the world of Agentic AI wherein intelligent systems autonomously plan, act, and learn to achieve complex business goals. Through a combination of lectures, demonstrations, and immersive hands-on labs, participants will design, build, deploy, and monitor autonomous multi-agent systems that integrate real-time reasoning, tool use, and memory. By the end of the course, participants will understand how to design AI systems that operate independently, integrate with enterprise tools, and continuously improve performance through MLOps-driven feedback loops.
Who should attend
- Application Developers
- Data Scientists
- Project Managers
- System Engineers
- Technology Trainers
Prerequisites
- Basic understanding of AI, familiarity with Python, and exposure to generative AI
Course Objectives
- Understand what agentic AI is, how it differs from generative AI, and its role in customer interactions
- Building and deploying an autonomous AI Agent
- Dive into the technical components and start building a simple agentic AI
- Focus on autonomy, planning, and scalability in agentic AI systems
- Explore practical applications and learn to deploy agentic AI in real-world settings
- Address ethical challenges, evaluate performance, and showcase projects
Outline: Agentic AI for Building Intelligent Customer Interactions (AA-BACIS)
Foundations of Autonomy and Tool Use
- Lecture: The Crisis of the Single Prompt: Why Agents?
- Lecture: Agentic AI vs. Static LLMs: A Conceptual Shift
- Lecture: Anatomy of an Agent: Planners, Memory, and Tools
- Lecture: Framework Comparison: LangChain vs. CrewAI
- Lecture + Lab: Set Up Environment and Agent Frameworks
- Lecture + Lab: Build a Basic “Hello World” Conversational Agent
Enabling Action: Tools and Functions
- Lecture: Designing Robust Tools and Interfaces
- Lecture: Safety Protocols: JSON and Pydantic Schemas
- Lecture + Lab: Define and Register Custom Python Tools
- Lecture + Lab: Implement Error Handling and Fallback Strategies
- Lecture + Lab: Build an Agent with Multi-Tool Logic (SQL + Email)
Agent Memory and State Management
- Lecture: Short-term vs. Long-term Memory Architectures
- Lecture: RAG and Vector Databases for Agents
- Lecture + Lab: Integrate Vector Store for Long-term Memory
- Lecture + Lab: Implement Persistent State Across Workflows
- Lecture + Lab: Optimize Retrieval for Agent Reasoning
Production, Orchestration, and MLOps Integration
- Lecture: Multi-Agent Patterns: Decomposition and Roles
- Lecture: Orchestrating Hand-offs with CrewAI
- Lecture + Lab: Define Specialized Agents (Researcher, Writer, Reviewer)
- Lecture + Lab: Build a Multi-Agent Crew for Complex Solving
- Lecture + Lab: Debugging Inter-Agent Communication Conflicts
Agent Reliability and Observability
- Lecture: The MLOps Crossover: Tracing Autonomy
- Lecture: Metrics for Agents: Success Rates and Latency
- Lecture + Lab: Instrument Agent Traces with ClearML
- Lecture + Lab: Version Control for Agent Configurations
- Lecture + Lab: Analyze “Chain of Thought” Logs for Errors
Production Deployment and Scaling
- Lecture: Agent Serving Architecture and Containerization
- Lecture: Cost Management and Token Optimization
- Lecture + Lab: Containerize the Agent as a Microservice
- Lecture + Lab: Optimize Inference Latency and Caching
- Lecture + Lab: Prepare Deployment Manifests for Scaling
Governance and Continuous Improvement
- Lecture: Detecting Agent Drift and Hallucinations
- Lecture: Security Guardrails and Ethical Governance
- Lecture + Lab: Implement Input/Output Guardrails
- Lecture + Lab: Simulate Drift and Trigger Retraining Loops