Securing AI workloads across On-Prem and Cloud
Join Cloud Native Tel Aviv at Imperva for an evening on securing AI workloads across on-prem and cloud: confidential GPU inference on Kubernetes, a WAF's journey to cloud-native, and securing coding agents with CNCF projects.
Agenda
18:00 - Gathering & networking
18:30 - Trust the Other: Confidential Computing for GenAI
19:00 - From Pod to Pop - A WAF's cloud-native journey. Human-built. AI-assisted.
19:30 - Securing Coding Agents - Across On-Prem and Cloud
Trust the Other: Confidential Computing for GenAI
Yuval Ostrovsky, Co-founder & CTO
How can you run AI inference on cloud GPUs without exposing model weights or sensitive prompts to the infrastructure operator?
In this session, we'll follow a confidential AI inference workload on Kubernetes, from deploying an encrypted model to serving requests inside a hardware-protected environment.
We'll explore how GPU confidential computing, remote attestation, and policy-controlled key release work together to protect models and data, and examine the trust boundaries between Kubernetes, the infrastructure operator, and the underlying hardware.
Designed for Kubernetes practitioners and platform engineers, this talk provides a practical understanding of how confidential GPU inference works, what it protects, and where its security guarantees end.
From Pod to Pop - A WAF's cloud-native journey. Human-built. AI-assisted.
Lital Hekmati, DevOps Engineer at Imperva-Thales
Vitaly Marom, Senior DevOps Engineer at Imperva-Thales
We took Imperva's WAF from physical PoPs in data centers to Kubernetes, Envoy and Argo CD inside AWS. This talk covers what CNCF gave a small team, why we chose Envoy, and what AI did and didn't do along the way. It made us faster, but understanding the system is still a human job.
Securing Coding Agents - Across On-Prem and Cloud
Adi Ziv, Solutions Architect, AWS
Ido Ziv, Senior Platform Engineer, Air
Coding agents now run everything with real credentials. The context they act upon can change after it got reviewed. In this session we will raise common issues and explore how to solve each of them and secure each step with your favorite CNCF projects: Envoy Gateway, OPA, and OTEL.
You'll leave with a deep understanding of how to solve real security issues of running agents across on-prem and cloud.
Organized by Or Navon
Contact organizers
-
6:00 PM - 6:30 PM IST
Gathering & networking
in-personArrive early for coffee and networking before the talks.
-
6:30 PM - 7:00 PM IST
Trust the Other: Confidential Computing for GenAI
in-personSpeaker: Yuval Ostrovsky, Co-founder & CTO
How can you run AI inference on cloud GPUs without exposing model weights or sensitive prompts to the infrastructure operator?
In this session, we'll follow a confidential AI inference workload on Kubernetes, from deploying an encrypted model to serving requests inside a hardware-protected environment.
We'll explore how GPU confidential computing, remote attestation, and policy-controlled key release work together to protect models and data, and examine the trust boundaries between Kubernetes, the infrastructure operator, and the underlying hardware.
Designed for Kubernetes practitioners and platform engineers, this talk provides a practical understanding of how confidential GPU inference works, what it protects, and where its security guarantees end.
SPEAKERS -
7:00 PM - 7:30 PM IST
From Pod to Pop - A WAF's cloud-native journey. Human-built. AI-assisted.
in-personSpeakers: Lital Hekmati, DevOps Engineer at Imperva-Thales; Vitaly Marom, Senior DevOps Engineer at Imperva-Thales
We took Imperva's WAF from physical PoPs in data centers to Kubernetes, Envoy and Argo CD inside AWS. This talk covers what CNCF gave a small team, why we chose Envoy, and what AI did and didn't do along the way. It made us faster, but understanding the system is still a human job.
SPEAKERS -
7:30 PM - 8:00 PM IST
Securing Coding Agents - Across On-Prem and Cloud
in-personSpeakers: Adi Ziv, Solutions Architect, AWS; Ido Ziv, Senior Platform Engineer, Air
Coding agents now run everything with real credentials. The context they act upon can change after it got reviewed. In this session we will raise common issues and explore how to solve each of them and secure each step with your favorite CNCF projects: Envoy Gateway, OPA, and OTEL.
You'll leave with a deep understanding of how to solve real security issues of running agents across on-prem and cloud.
SPEAKERS
No sessions match the selected labels.