04-801-W3 Agentic AI: Foundations and Applications
Location: Africa
Units: 6
Semester Offered: Spring
Course description
Agentic AI-AI systems that act autonomously, plan strategically, and use tools dynamically-is redefining how we build software, automate work, and design intelligent interfaces. From AutoGPT to enterprise-grade AI agents, the last two years have seen an explosion in systems that turn language models into reasoning, planning, and decision-making agents. This mini-course gives students a rapid, hands-on immersion into this emerging paradigm.
In just six weeks, students will learn how to architect goal-directed agents powered by large language models (LLMs), connect them to tools, APIs, and memory systems, and orchestrate them for multi-step real-world tasks. Through demos, implementation sprints, and critical case studies, students will explore cutting-edge frameworks like ReAct, LangChain, and AutoGen. They’ll also tackle agent-specific challenges like hallucination control, self-evaluation, multi-agent collaboration, and system integration.
The course is ideal for students who want to move beyond prompt engineering and start building robust, interactive agents that automate reasoning and execution in domains like business intelligence, research assistance, policy analysis, and technical operations. Each session is anchored in practical exercises and architectural thinking-designed to build expertise for designing agentic systems, not just using them.
Learning objectives
- Understand key concepts behind agentic AI and intelligent LLM agents
- Build agents that can use tools, reason step-by-step, and plan over multiple turns
- Gain fluency with core patterns and frameworks such as ReAct, LangChain, and AutoGen
- Critically evaluate agent behavior, limitations, and implications for deployment
Outcomes
After completing this course, students will be able to do the following:
- Complete in-class coding demonstrations and agent-building exercises
- Do a short team project or use-case writeup describing an implemented agentic system
- Presentate and have critical discussion of architecture, design rationale, and ethical impact
Content details
Week 1: Introduction to Agentic AI
- What are AI agents? Why now?
- Evolution from prompt engineering to agentic architectures
- Live demo of a modern agent (e.g., AutoGPT or LangChain tool agent)
Week 2: Reasoning and Planning with LLMs
- Chain-of-Thought prompting and step-by-step reasoning
- Architectures for long-horizon thinking
- Exercises: Writing and testing structured prompts
Week 3: Tool Use and Retrieval-Augmented Agents
- Connecting agents to calculators, search APIs, and custom tools
- Memory and vector databases (e.g., FAISS, Chroma)
- Frameworks: Introduction to LangChain agent interfaces
Week 4: Multi-Agent Orchestration and Collaboration
- Planning with multiple LLM agents (AutoGen, CrewAI)
- Role specialization, task routing, and coordination patterns
- Group activity: Design a collaborative multi-agent system
Week 5: Evaluation, Testing, and Agent Debugging
- How do we measure agent performance?
- Benchmarks, hallucination mitigation, and reflection loops
- Tools: AgentBench, OpenAI evals, manual scenario testing
Week 6: Use Cases and Domain Applications
- Business process automation, research workflows, and policy agents
- Guest case studies or paper reviews (e.g., Microsoft AutoGen, ReAct paper)
- Teams finalize their mini-projects
Week 7: Project Presentations and Futures
- Team presentations: architecture, demo, and evaluation
- Peer feedback and instructor critique
- Final reflections on societal impact and future directions of Agentic AI