Join Antoine Delignat-Lavaud, researcher at Microsoft, as he dives into the technical and practical challenges of building confidential inferencing and SaaS on Azure. This session explores the evolution of confidential computing, highlighting different levels of guarantees, from isolated services to fully transparent, auditable confidential AI.
Learn about the complexities of scalable confidential services, secure data transport, attestation, and key management using protocols like Oblivious HTTP. Discover how stateless load balancing, modular key release policies, and Azure confidential ledger enable robust auditability and transparency, even in dynamic Kubernetes environments.
This talk also covers integration with cloud transparency services, confidential build pipelines, and rolling out code updates while maintaining full auditability. Experience a live demo of confidential inferencing, including speech-to-text and multi-GPU model deployments, and get insights into secure server-to-server communication and transparent attestation via DNS.
For other Confidential Computing Summit 2025 sessions visit: https://msft.it/6056t3Uj2
Learn more about Azure Confidential Computing by visiting: https://msft.it/6057t3UjN
