Programme Content
What is Agentic AI: concepts, scope, and positioning in the AI landscape
Autonomy, goals, actions, and feedback loops in agent systems
High-level agent architecture: observe \(\rightarrow \) reason \(\rightarrow \) act \(\rightarrow \) learn cycle
Agentic AI vs GenAI, automation, and traditional AI systems
Model Context Protocol (MCP): context management, state handling, and tool contracts
Real-world use cases and industry applicability
Module-end Project 1: Agentic System Design Blueprint
2. LLMs as Reasoning Engines for Agentic SystemsRole of LLMs in agentic decision-making and autonomous reasoning
Reasoning workflows and structured prompting techniques
Advanced prompt engineering and prompt patterns (ReAct, reflection, constraints)
Planning, task decomposition, and self-reflection mechanisms
Reliability, constraints, and failure modes of LLM-based agents
Designing dependable reasoning pipelines for production systems
Cyber-physical systems integration: real-world case studies
Module-end Project 2: Autonomous Prompt Generation & Refinement System
3. Designing Autonomous Multi-Agent SystemsSingle-agent vs multi-agent systems: architectural differences
Role-based agents and dynamic task distribution
Communication protocols and coordination mechanisms between agents
System design principles, scalability considerations, and complexity trade-offs
Real-world multi-agent use cases: enterprise workflows, collaborative problem-solving
Module-end Project 3: Multi-Agent Collaboration System
4. Agentic Workflows, Automation & Decision OrchestrationEnd-to-end agentic workflows: from goal specification to execution
MCP-based orchestration of agents, tools, and external APIs
Enterprise automation and integration with legacy systems
Monitoring, control mechanisms, and human-in-the-loop oversight
Event-driven architectures and real-time agent responsiveness
Module-end Project 4: Agent-Ororchestrated Automation Pipeline
5. Real-World Agentic AI Engineering (RAG, Deployment, Governance)Retrieval-Augmented Generation (RAG) for agentic systems
Long-term agent memory architectures and state management
Deployment considerations: cloud, edge, and hybrid architectures
Governance frameworks, risk management, and responsible autonomy
Evaluation metrics, continuous monitoring, and compliance requirements
Enterprise and industry perspectives on agentic AI adoption
Designing and Deploying an Enterprise-Ready Agentic AI System
Objective: Design, build, and evaluate an end-to-end agentic AI system that solves a real-world problem, demonstrating autonomy, reasoning, orchestration, and responsible deployment. The system should be capable of observing its environment, reasoning through complex scenarios, taking autonomous actions, and learning from outcomes, all while operating reliably in a simulated enterprise context.
Project Scope
Customer Support Automation: Intelligent agent handling complex multi-turn conversations with knowledge retrieval.
Enterprise Workflow Automation: Multi-agent system coordinating across departments (procurement, approvals, notifications).
Research & Analysis Agent: Autonomous system gathering, synthesizing, and reporting on specific domains.
Cyber-Physical System Control: Agent-based monitoring and decision-making for IoT/sensor network.
Custom Domain: Student-proposed application with instructor approval.
Note: The list of projects and tools are indicative and can be modified at the discretion of the Programme Coordinator.
