Today, I spent the day advancing my preparation for AI-102: Designing and Implementing Microsoft Azure AI Solutions. I realized my Tutorial Dojo and Whizlabs practice courses didn't cover Semantic Kernel, Autogen and Azure AI Foundry orchestration in as much depth as needed, so I went into it myself:
- Explored Semantic Kernel (SK) for advanced prompt engineering and lightweight orchestration.
- Dove into Azure AI Foundry Orchestration for building multi-agent workflows with built-in governance and scalability.
- Learned about Autogen and key agent types (Planner, Retriever, Writer) and their specialized roles.
- Clarified Bot vs. Agent: Bots for simple, scripted interactions; Agents for dynamic, goal-oriented tasks with reasoning and tool use.
Key takeaway:
Azure AI Foundry shines for enterprise-scale, distributed multi-agent systems with strong governance, observability, and compliance. Semantic Kernel is ideal for simpler, flexible setups and it can even be containerized as microservices on AKS for better cost control and portability.
Conclusion: Use Semantic Kernel for prototyping/simple projects; Azure AI Foundry for complex, production-grade agentic workflows.
Next up: Building a small multi-agent project with Semantic Kernel orchestration and an Ollama container for local LLM testing!