Beyond the Hype: Real-World MLOps, GPU Bursting & FOSS Security

Join us in Vienna for an enlightening evening where DevOps meets Security! Our Vienna DevOps and Security Meetup is the perfect place to connect with fellow tech enthusiasts, industry experts, and thought leaders. Whether you're a seasoned professional or just starting, this event is designed to provide valuable insights and networking opportunities.

This meetup will take place at ÖBB Open Innovation Factory at Lassallestraße 5, Foyer 1, 1020 Wien. Thanks to ÖBB for sponsoring the location!

Agenda:

• 18:00 - 18:30: Arrival and Networking
• 18:30 - 19:00: Talk 1
• 19:00 - 19:30: Break and Refreshments
• 19:30 - 20:00: Talk 2
• 20:00 - 20:15: Break and Refreshments
• 20:15 - 20:45: Talk 3
• 20:45 - : Networking

Sessions:

Let the Bot Review it: Reducing Toil with Fossabot - Julian Wei Zhuang (WhizUs)
Dependency update PRs look innocent but can hide critical security vulnerabilities and breaking changes—even in patch bumps. Discover how AI tools like Fossabot and Claude can automatically catch what manual reviewers miss, from npm packages to Helm charts, and learn best practices for combining AI analysis with human judgment.

The Hybrid AI Stack: Rancher, Envoy, KubeAI and the Art of GPU Bursting - Elias Rami (WhizUs)
This talk presents a hybrid AI infrastructure stack for self-hosting LLMs on Kubernetes. It covers the evolution from a simple Ollama setup to a production-grade architecture using Rancher RKE2 with bare-metal GPUs (H100s with MIG partitioning), KubeAI for Kubernetes-native model serving via vLLM, and Envoy AI Gateway with the Gateway API Inference Extension for intelligent, KV-cache-aware routing.

ML(Dev)Ops at Mercari - Norbert Preining (Mercari)
Mercari is the biggest client-to-client marketplace in Japan with more than 23M monthly active users (1 in 4 adults in Japan), and over 3B listings to date.
In the last years, we have introduced reranking, a ML based system to improve search results, into the search system.
We will talk about the slow and incremental improvements in our ML/DevOps setup from an incredible simple in-search adapter model
to an elaborated setup with multiple models, all while running
under full load with latency budgets in the tens of miliseconds.

RSVP: Don't miss out on this opportunity to elevate your skills and network with industry leaders. Seats are limited, so secure your spot today!