Module 17: Azure AI Foundry Local¶
Level: L300 | Duration: 3 hours | Prerequisites: Module 14 | Hands-on: Lab
Module Status
Framework module — content scope under research. See research/ai-foundry-local.md.
Learning Objectives (tentative)¶
- Describe Azure AI Foundry Local and its value for sovereign / low-latency AI workloads
- Deploy Foundry Local on Azure Local infrastructure
- Deploy small language models (Phi, Llama variants) locally
- Configure GPU partitioning and SR-IOV on Azure Local for AI workloads
- Call local inference endpoints from workloads on the cluster
Topics (tentative)¶
- Azure AI Foundry vs. Foundry Local — cloud vs. on-premises
- Use Cases — sovereignty, data residency, low-latency inference
- Hardware Requirements — GPU SKUs, memory
- GPU Partitioning and GPU SR-IOV on Azure Local
- Deployment on AKS Arc — Foundry Local containers
- Model Management — pulling, deploying, updating models
- Inference Endpoints — calling local APIs
- Security and Compliance Considerations — keeping data on-prem
Hands-on¶
Lab: Deploy Foundry Local on AKS Arc. Deploy a Phi-3 (or equivalent) small model. Call inference endpoint from a test app. Validate no data egress.
IaC: Bicep templates in labs/iac/17-ai-foundry-local/ (planned).
Related Resources¶
- Azure AI Foundry Local
- Research notes: research/ai-foundry-local.md
- Slides:
slides/17-ai-foundry-local/(planned)