Agentic AI
Visiting Lecturer
Visiting from Herald College Kathmandu (University of Wolverhampton)
Course Description
An applied introduction to agentic AI for first-year students, assuming no programming background. The course moves from how language models actually generate text through to designing, deploying and governing systems of cooperating agents. Every module pairs a lecture with a hands-on lab, so students build the thing being described in the same week they hear about it — a working assistant, an automated workflow, a multi-agent system — and finish with a capstone of their own design.
Learning Outcomes
By the end of the course, students can:
- Explain how language models generate text and how decoding parameters shape output.
- Write structured, reusable prompts and build a custom AI assistant.
- Design an agent with explicit planning, memory and tool use, and build one without code.
- Automate a multi-step business workflow with n8n and connected data sources.
- Compose several agents into a system, and argue for its reliability and governance.
Schedule
AI Foundations
Introduction to how modern AI systems work and why the same prompt produces different results each time. Students ran structured experiments on temperature, top-p and output length to see decoding parameters change behaviour first-hand, then mapped the current AI landscape and how to choose the right tool for a task.
Prompt Fundamentals & Advanced Prompting
Moving from vague requests to reusable AI assistants. The first session covered prompt structure, context and constraints; the second built on that with advanced techniques for turning good prompts into consistently great results.
Designing, Building & Governing AI Agents
The shift from chatbot to collaborator — how agents plan, remember and act. Students built a working agent without writing a line of code, and looked at what governing one responsibly requires.
AI Workspaces & Connected AI
Building a persistent AI workspace using projects, skills, plugins and artifacts, then connecting it to external data for document intelligence. Session also covered the Docker and n8n setup students would need the following week.
AI Automation with n8n
Building automated, AI-powered workflows with Docker and n8n. Concepts were introduced and then built live, culminating in an end-to-end student report workflow driven from a sample dataset.
AI-Assisted Coding
Turning natural-language ideas into working software using AI coding tools. Students set up their environments, then spent the second session ideating, building, refining and presenting a project of their own.
Autonomous & Multi-Agent Systems
How agents decide, act, remember and stay reliable without a human driving every step, then scaling from one agent to a team of agents running a whole workflow. Closed with agent governance — deploying safely and demonstrating business value.
Capstone Projects
Students present the agentic system they designed and built across the course, with peer and instructor critique on design decisions, reliability and governance.