Secure Multi-tenant AI is a security-focused session on building shared AI platforms without compromising data or compliance. The course opens by examining the core challenge of multi-tenancy: sharing GPU hardware, storage, and network paths across customers or teams. This sharing creates opportunities for data leakage and model IP theft.
Students then work through the security concerns unique to multi-tenant environments. Topics include workload isolation with containers, VMs, and confidential computing, identity propagation across services and agents, secrets management, and auditability. Labs and demos show secure patterns for enterprise deployments, giving data scientists, data engineers, and MLOps professionals a shared vocabulary for evaluating platform risk.
The course also surveys the software and hardware solutions available for multi-tenancy, comparing isolation strategies and their tradeoffs in performance and security. A closing module looks at hardware implementations for multi-tenant GPUs. It covers how contention for GPU resources drives latency and service degradation, and what partitioning and scheduling approaches keep tenants isolated. Attendees also address the regulatory compliance concerns that shared AI platforms raise for data residency, auditability, and access control. Upon completion, attendees will have a solid understanding of the concerns around multitenant AI security and the evolving tools available for addressing them.
Who Should Attend
Data Scientists, Data Engineers, Machine Learning Engineers, MLOps professionals
What Attendees Will Learn
Upon completing Secure Multi-tenant AI, participants will be able to:
- Explain the challenges of building AI applications for multiple tenants
- Identify how GPU hardware contention leads to latency and service degradation
- Apply model IP protection techniques when hosting multiple models on shared hardware
- Address regulatory compliance concerns in multi-tenant AI platforms
Prerequisites
Participants should have basic programming skills (preferably in Python), a strong understanding of mathematics and statistics, intermediate familiarity with machine learning, a grasp of computer science fundamentals, familiarity with common operating systems and basic command-line operations, and a willingness to review pre-course materials.
These prerequisites will help participants engage effectively with the course material and hands-on labs, making the learning experience more rewarding.