Certificate Program in Agentic AI

This week, learners will be introduced to the foundational concepts of Generative AI and Large Language Models. They will gain an understanding of how LLMs are trained, the architecture that powers them, and the principles that guide their generative behavior. The week will also explore the evolution of prompting methods: Zero-Shot, One-Shot, and Few-Shot, and their role in guiding model outputs. Learners will examine real-world use cases across business functions such as Marketing, Finance, Healthcare, and Productivity, enabling them to recognize the growing influence of Generative AI in industry applications.

This week, learners will explore advanced Prompt Engineering strategies to enhance the effectiveness of natural language interactions with LLMs. They will delve into Chain-of-Thought Prompting, which improves reasoning by guiding the model through intermediate steps. The week will also introduce the concept of Retrieval-Augmented Generation (RAG), which combines the generative capabilities of LLMs with the precision of external data sources. Learners will gain hands-on experience working with vector databases and understand how integrated retrieval systems improve the contextual accuracy of AI-generated outputs.

This week, learners will be introduced to the foundational concepts of Agentic AI, distinguishing it from traditional AI systems. The week focuses on the characteristics that define agentic behavior, including autonomy, perception, and goal-driven actions. Learners will understand the concept of rational agents as models for autonomous decision-making and discover how agentic workflows orchestrate the interaction between perception, reasoning, and action in dynamic environments.

This week, learners will explore the essential properties of intelligent agents and the types of environments in which they operate. The week introduces the PEAS framework (Performance, Environment, Actuators, Sensors) to define the task environment of an intelligent agent, outlining how it perceives and acts in the world. Learners will compare various types of agents by assessing their levels of autonomy and goal complexity to understand how different configurations impact the behavior of AI agents.

This week, learners will gain insights into how agents plan and reason through different paradigms. The week covers classical planning methods as well as reasoning approaches powered by LLMs, including Chain-of-Thought (CoT) and ReAct (Reasoning and Acting). By exploring these mechanisms, learners will learn how agents deconstruct complex tasks and take rational actions to achieve their goals.

This week introduces learners to symbolic reasoning frameworks, specifically the Belief-Desire-Intention (BDI) model, and examines how these symbolic structures underpin agent behavior. Learners will also explore Neuro-Symbolic AI (NSAI), where the reasoning power of symbolic models is enhanced through the learning capabilities of neural networks. This hybrid approach enables the design of agents with robust decision-making skills in complex environments.

This week, learners will examine the core modules of LLM-based agents: memory, planning, and action, and understand how they work together to structure intelligent behavior. The concept of scaffolding will be introduced to demonstrate how agents manage workflows, while the use of external tools through action modules will be explored to understand how agents extend their functionality.

This project focuses on developing an intelligent agent to automate the processing of employee-submitted expense bills. The agent will analyze submissions and generate expense claim documents that are fully compliant with established company expense policies. Domain: Productivity and Workflow Automation.

This week focuses on how agents can improve decision-making through reinforcement learning (RL). Learners will explore both foundational RL methods and advanced techniques like Deep RL and Verbal RL. The week emphasizes how learning from rewards enables agents to develop optimal behaviors and adapt to new environments, laying the groundwork for long-term agent adaptability and lifelong learning.

This week, learners will explore the dynamics of Multi-Agent Systems (MAS) where multiple intelligent agents interact, coordinate, and communicate. Core concepts like coordination, cooperation, and basic communication protocols will be covered, alongside game theory principles for analyzing strategic interactions. The integration of Large Language Models into MAS architectures will also be explored to show how agents can reason and communicate in more natural and effective ways.

This week introduces learners to Human-Agent Interaction (HAI) and the design of embodied agents that can operate in real-world or simulated physical spaces. Topics include establishing trust, creating common ground for interaction, and the use of virtual environments in training embodied agents. Learners will understand the importance of emotional intelligence, empathy, and contextual awareness in designing agents that support human needs.

This project involves creating an automated research agent designed to retrieve and synthesize information from multiple data sources, including the internet. The agent will provide comprehensive and iteratively refined answers to user queries, enhancing the depth and accuracy of information retrieval.

This week introduces learners to the AI alignment problem and the key ethical challenges specific to agentic AI systems. Topics include safety concerns, responsibility in autonomous agents, and frameworks for responsible AI governance. Learners will analyze unique risks posed by agentic properties and apply ethical reasoning to assess agent scenarios, ensuring safe and trustworthy AI deployment.

This week covers the technical landscape of agentic AI development frameworks. Learners will explore popular tools such as LangChain, AutoGen, and CrewAI for building single-agent Retrieval-Augmented Generation (RAG) systems, as well as multi-agent reinforcement learning frameworks like PettingZoo. The focus will be on selecting appropriate frameworks for different agentic tasks and understanding how to effectively coordinate multi-agent environments.

This project entails the development of a customer support chatbot engineered to address frequently asked questions (FAQs) concerning a specific product or service. The agent will leverage a predefined knowledge base, employ advanced knowledge retrieval techniques, engage in natural language interactions, and integrate a Large Language Model (LLM) to enhance communication clarity and user experience.

Domain: Customer Support & Automation