Cloud Native London, May 2026

Hi folks!

Welcome to our May Cloud Native London meetup! Join us to hear from our three great speakers and network with your fellow techies over pizza and drinks, or alternatively chat and following along on Youtube or LinkedIn!

6:00 Pizza and drinks
6:30 Welcome
6:45 The End of Handoffs: Build, Run, Own in One Flow (Alberto Pose, Pulumi)
7:15 You Can’t Patch Fast Enough: What AI-Driven Attacks Mean (Idan Elor, Oligo Security)
7:45 Break
8:00 From Zero to HyperPod: Distributed Model Training on AWS (Anton Nazaruk, Cloud Combinator)
8:30 Wrap up

See you there!

Cheryl (@oicheryl)

The End of Handoffs: Build, Run, Own in One Flow (Alberto Pose, Pulumi)
AI is changing the shape of team structures. A similar shift to the one that brought dev and ops together into DevOps is now happening as product must also meet the accountability bar.
At the same time, tooling has not caught up with "throwing it to Claude/Codex". Engineers are confronted with a choice: trust code and actions they do not fully understand, or... move too slowly.
Are these problems fundamentally new, or are there lessons we can draw from the past to adapt and think about what comes next?

Alberto Pose is a software engineer with a soft spot for developer tooling and infrastructure. He is currently part of the team managing the CI/CD pipelines for Pulumi's open source projects. Before this, he spent nearly ten years at Prime Video and AWS. A major highlight of his time there was helping bootstrap the living room device automation team, taking it from a small group effort to a 30 person organisation that brought full automation to millions of streaming devices worldwide.

You Can’t Patch Fast Enough: What AI-Driven Attacks Mean (Idan Elor, Oligo Security)
Anthropic’s Project Glasswing announcement put the industry on notice. AI has fundamentally changed the threat landscape and the risks facing organizations, and it’s happening faster than most security models can adapt.
AI-assisted attackers can now discover vulnerabilities, generate exploits, and launch attacks at rates never seen before, collapsing the window between disclosure and exploitation. In this environment, patching alone is no longer enough.
For cloud-native applications, where systems are dynamic and constantly evolving, this creates a new challenge: how do you stay protected when you can’t fix vulnerabilities fast enough?
This talk explores how the attacker model is shifting in the AI era, why traditional approaches are breaking down, and what it means to move from vulnerability-based security to real-time protection at runtime. We’ll cover how focusing on exploit techniques enables teams to stop attacks as they happen, including zero-days.

Idan Elor is Field CTO at Oligo Security, where he partners with large enterprises to solve complex application and cloud security challenges. He most recently served as Director of Solution Engineering & Tech-Alliances at Apiiro, where he empowered enterprises to secure their software supply chains. With over a decade of experience spanning application security, DevSecOps, and mobile security, Idan has also held leadership positions at companies like Snyk, Symantec, and HP. His unique background combines deep hands-on technical expertise. A passionate advocate for bridging the gap between security and development teams, Idan is known for his ability to translate complex security concepts into actionable strategies that organizations can actually implement.

From Zero to HyperPod: Distributed Model Training on AWS (Anton Nazaruk, Cloud Combinator)
You've got a model that works. You just need more GPUs. How hard can it be? That's where the pain starts.
GPU availability, infrastructure complexity, and cost are the three blockers that trip up even experienced teams when scaling from single-GPU training to serious distributed workloads.
This talk is a practical walkthrough of how to set up distributed model training on AWS - covering the capacity options (On-Demand, Spot, Capacity Blocks, SageMaker Training Plans), when to use each, and a repeatable infrastructure blueprint for compute, networking, storage, and observability. I'll demo provisioning a HyperPod cluster and running a distributed training job with automatic failure recovery, and share the cost levers that matter at scale.
Whether you're a platform engineer supporting ML teams or an ML engineer tired of fighting infrastructure, you'll leave with a decision framework and a blueprint you can implement.

Anton Nazaruk is CTO at Cloud Combinator, where he helps companies run GPU workloads on AWS - from early-stage startups to larger organisations doing distributed training at scale. He focuses on making ML infrastructure repeatable, resilient, and cost-efficient.
*LinkedIn: linkedin.com/in/anton-nazaruk*

Check out https://www.oicheryl.com/cloudnativelondon if you're interested in speaking or sponsoring.