Multi AI Agents for Content Creation

Getting several AI agents to plan, retrieve information and act together is a problem Dr. Alfred Ang solves in day-to-day engineering, not just in the classroom. Dr. Ang is a full-stack AI developer and Agentic AI solution architect, a corporate trainer, and Chief Technology Officer of Tertiary Infotech Academy Pte Ltd in Singapore, the organisation behind Tertiary Courses. A good example is an automated video production pipeline in which a topic request from a web form triggers an agent that researches the subject, writes the script, renders the animation, verifies the result and publishes it, each stage handing work to the next. Dr. Ang has also built retrieval-augmented chatbots with n8n, including a Telegram agent that answers questions from uploaded PDFs and a customer-service assistant whose course brochures are ingested into Supabase, Pinecone or Qdrant vector databases. Further projects include an ERP prototype where AI agents prepare purchase orders that pass through human-in-the-loop approvals by email, Asana or Telegram, a stock-analysis Telegram agent that reasons through Buy, Sell or Hold recommendations, and an AI-powered content platform built on the Claude Agent SDK with a self-improving chatbot. Beyond agents, Dr. Ang builds complete products end to end, from front end and back end to databases, cloud deployment and native mobile apps, including the self-hosted AI learning and training management system used to run classes. More than 300 public repositories document this work on GitHub, offering a rare look at how multi-agent ideas behave once real users and real data are involved.

In this Multi Agents System with AutoGen course, Dr. Ang uses that experience to keep the focus on agents that actually finish useful work. The course begins with the foundations: how large language models behave, what turns a model into an agent, and how AutoGen compares with CrewAI as a framework for building them. Learners then create a first agent with AutoGen AgentChat, giving it a clear role, instructions and tools, and watching how it reasons through a task. The second topic tackles retrieval-augmented generation, explaining why agents need access to fresh, trusted information and guiding learners through building a RAG agent that grounds its answers in their own documents rather than guessing. The final topic brings everything together in multi-agent workflows, where learners design agents with distinct responsibilities, decide how they hand work to one another and orchestrate the whole team in AutoGen to automate a complex task. Dr. Ang teaches by building in small steps, running each version, reading the agent conversation and adjusting roles or prompts when the output drifts, which mirrors how agent systems are refined in practice. Lessons drawn from production work run throughout: keeping each agent's job narrow, inserting a human review point where decisions matter, and testing behaviour before trusting it with real tasks. Learners leave with working AutoGen code they can extend for their own organisation and a clear mental model of when multiple agents add value. For more examples of agentic systems in action, browse Dr. Ang's public repositories at https://github.com/alfredang