BRK370 Breakout Tue 17 Nov 16:30 PST
Build integrated AI apps with Azure HorizonDB
Jonathon Frost, Brian Holt, Dylan Addington
Session description
Microsoft's catalogue description
Build AI applications grounded in operational data without assembling a fragmented data stack. See how Azure HorizonDB brings PostgreSQL data together with AI functions, durable AI pipelines, vector and hybrid search, semantic ranking, and knowledge graphs. Through a fast-paced technical walkthrough and live demo, learn how to transform continuously changing operational data into powerful, connected, and context-aware AI experiences.
Demo timestamps
Microsoft's player does not support links to a specific time, so open the recording and skip to the time shown.
- Skip to 0:02:39
AI pipeline over manufacturer tables
In VS Code, an Azure AI extension function creates a pipeline with an entity-extraction step and an embed step over manufacturer profile and raw content tables with an empty AI-context column.
- Skip to 0:08:53
Vector, full-text and hybrid search
The same search is run as vector search, then full-text (top hit changes), then hybrid search using reciprocal rank fusion so results share a single score.
- Skip to 0:11:24
Knowledge-graph primitives with Apache AGE
A graph of suppliers, plants, certifications and inspection findings is shown, built with a HorizonDB knowledge graphs for AI pipelines feature described as reducing the code needed.
- Skip to 0:15:31
End-to-end pipeline when a new finding arrives
Hybrid search returns suppliers B and A. A new profile version lands from a legacy system, and the pipeline is run so the new finding can change who qualifies.
- Skip to 0:26:23
Pre-recorded healthcare skills-catalogue application
A pre-recorded HorizonDB app takes a prompt to register a patient, add a guarantor and book a provider. The skills catalogue is searched, a plan is shown for human approval, then execution continues.