On NOV 6, I presented at a Memphis technology event, SIM Memphis Strategy Series for Executives. My session was titled “A Journey to AI RAG with Semantic Kernel”, where I attempted to show how a .NET developer could leverage AI and build up to RAG scenarios.

The Event

The SIM Memphis Strategy Series is the largest gathering of IT leaders in the Memphis region each year. The event offers IT leaders the opportunity to educate one another in an environment where ideas are exchanged by attendees as well as outstanding speakers throughout the day’s agenda. I highly recommend getting this day on your calendar, as there were IT leaders present from around town.

However, the group behind this is SIM Memphis. To stay connected to these leaders and hear the pulse of IT in this region, I suggest connecting with this group - as well as attending their other meetings through the year. They cover a variety of topics. As well, they do way more than just have meetings. BUT, the meetings I have attended are educational.

Thanks to Peter Kim and Jim Howard for the presentation invite and event logistics. A shout out to Andrew Wylie for helping me get in the back door of the Memphis Hilton during a rain storm.

My Presentation

Back to my presentation… The day after a hot presedential election is an interesting day to present on AI and Semantic Kernel.

The outline was as follows:

• AI Starter
  • a quick runthrough of a few AI terms
  • demo: Simple Demo, where we send some simple questions to a GPT-40-Mini LLM, and see that it’s smart but lacks current knowledge (like current date, news events, and the strawberry dilemma)
• Azure AI OpenAI
  • OpenAI models hosted by Azure
  • cost considerations
  • demo: Wordle AI player
• Semantic Kernel
  • a fast intro to the open-source SDK, and how it works with AI capabilities (OpenAI and Azure are not required, but that’s what I demo’d with)
  • Semantic Kernel’s plugin capability (native and OpenAPI) --> powerful !
  • demo: Lights, where we shows lights controlled with the help of SK and AI
• RAG
  • overview and related terms, e.g. vector, proximity, embeddings, etc.
  • a look at the architecture of RAG solution, using Azure OpenAI along with Cosmos DB for NoSQL
  • demo: Product Search, showing vector storage, querying via Cosmos Db vector querying capability, semantic cache, etc.
• Agents
  • a fast look at agents in SK, not to be confused with “agentic patterns”
  • demo: multi-agent / multi-turn scenario [demo here]

Thanks to those who attended my session.

A Few Side Notes

• there was a lot of AI buzz at this event. It’s evident that AI is a big deal and the market is trying to figure it out.
• I recommend you to take a look at the AI app templates, which are promoted as “Build AI applications with pre-made templates”. When doing the Products demo, this was an example explained by this article, but they had a template that would build out the Azure infrastructure that drives the scenario. Well done !
• SK’s Process Framework could be very interesting, and I could not fit that topic into this presentation.