Kubernetes Platform Engineering - SF Bay Area

RSVP here: https://luma.com/eavisvo3
(Only 110 seats, limited seats left)

Due to the overwhelming community response and interest, we are squeezing in 6 talks, 15 minutes each.

​Agenda:

• 5:30 - 6:00: Welcome and Registration
• 6:00 - 6:45: Talks 1 -3 (details below)
• 6:45 - 7:15: Break & Networking
• 7:15 - 8:00: Talks 3 -6 (details below)
• 8:00 - 8:30: Wrap-up and Networking

Talks:

1. Distributed LLM Inference at Scale (Nutanix): As LLMs continue to grow in size, efficient distributed inference has become essential for production deployments. This talk covers the key architectural principles
2. Zero-Trust by Design: Securing Distributed Data Platforms in the AI Era (StitchFix): This talk outlines how to retrofit zero-trust security into complex, distributed data platforms that support modern AI workloads. Instead of traditional perimeter-based security, the approach utilizes identity federation, token-based authentication, and context-aware, fine-grained authorization (ABAC).
3. Introducing OttoFlow: Deterministic AI Workflows for Kubernetes (Nirmata): Most AI projects fail due to unpredictable reasoning, runaway costs, and security risks. OttoFlow addressed these core issues by shifting from open-ended prompts to deterministic, Kubernetes-native workflows that prioritize strict governance. It turns AI from a risky experiment into a reliable, production-grade tool for complex cluster operations. Learn how you can use OttoFlow, a new agentic workflow framework from the creators of Kyverno, to address core platform engineering challenges.
4. Building AI-Native Testing Platforms with Autonomous Cognitive Navigation (Uber): This session introduces Autonomous Cognitive Navigation (ACN), an AI-native framework designed to overcome the maintenance and scalability limitations of traditional, script-based testing in modern software platforms. By leveraging Large Language Models and state-aware reasoning, ACN enables intent-based automation that adapts to evolving user interfaces and validates workflows through natural language instructions. The talk outlines key engineering principles, including persistent memory and multimodal verification, that allow teams to integrate intelligent testing into CI/CD pipelines. Real-world implementation results demonstrate that this approach can achieve a 95% reliability rate while reducing the time required to create new test flows from 120 hours to under 4 hours.
5. Taming the Data Beast: AI-Powered Cardinality Control for Platform Engineering (NVIDIA): High cardinality in observability data is a major operational challenge for distributed systems, often causing inflated storage costs and slower incident response. This talk explains how platform teams can combat cardinality explosion by building scalable "golden paths" that utilize AI-driven anomaly detection and intelligent pre-aggregation. Attendees will receive a practical playbook for improving resilience, including strategies for label design, AI-assisted sampling, and tiered retention policies. Ultimately, the session provides actionable methods for platform engineers and SREs to reduce costs and maintain system stability.
6. To be announced:

​Interested in submitting a talk for the next session? Submit your proposals here: https://sessionize.com/platform-engineering1/

To reserve your spot RSVP at: https://luma.com/eavisvo3