From open weights to workload: Operationalizing at enterprise scale | BRK246
Pavan Tagirisa, Miguel Armenta, Manoj Bableshwar, and Gulsimo Osimi discuss the infrastructure and operational decisions needed to take open-weight models into production, with Microsoft Foundry Managed Compute and AT&T’s OTel 2.0 experience as the focus.
From model selection to enterprise deployment
Open-weight models give teams greater choice and control, but deploying them at enterprise scale requires more than selecting a model. This session examines the infrastructure, serving stack, and operations needed to support production workloads, and how Microsoft Foundry Managed Compute helps teams move from selection to deployment.
AT&T’s OTel 2.0 experience provides the example of using multiple open models and GPU types rather than standardizing on a single model. The discussion centers on balancing three considerations:
- Performance: Matching model and serving infrastructure to workload requirements.
- Economics: Considering costs alongside performance when choosing models and GPU types.
- Workload fit: Supporting different needs through a multi-model approach.
Session details and resources
- Event: Microsoft Ignite 2026
- Session: BRK246 | Agent Platform
- Format: Breakout
- Level: 200 — Intermediate
- Resource: Agent Platform next steps