From the course: Microsoft Azure AI Engineer Associate (AI-102) Cert Prep by Microsoft Press
Join today to access over 25,100 courses taught by industry experts.
Trace, collect feedback, and reflect models - Azure Tutorial
From the course: Microsoft Azure AI Engineer Associate (AI-102) Cert Prep by Microsoft Press
Trace, collect feedback, and reflect models
- [Narrator] How do we make our LLMs better? Well, what do you mean by better? (laughing) Azure AI Foundry, as we saw, is a unifying web experience that has monitoring as one of its modules. That speaks to traceability. Another would be, it seems to me, integration with Git source control. A core part of DevOps is short feedback loops. I had mentioned including in your solutions things like user, thumbs up, thumbs down. Those little things can make a big difference to actually ingest that data and let it teach you. We've got lower-cost options, like Cosmos DB even has a free tier where we can persist things like log feedback. We've got App Insights, which is a platform service that I'll introduce you to in the demo. And then in Foundry there's this Prompt Flow architecture that's analogous in some ways to LangChain, if you're familiar with that, where we can integrate tracing observability and feedback. For example, Northwind Traders logs Q&A thumbs-down feedback and uses Prompt Flow…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.