THE THREE DISCIPLINES OF CI/CD SECURITY // DANIEL KRIVELEVICH DevOpsDays Tel Aviv 2021-12-16 | CI/CD pipelines are quickly becoming the path of least resistance for would-be attackers into sensitive internal systems, gaining access to critical data, with minimal effort.In the InfoSec world when we talk about CI/CD security often times this focuses on specific aspects of securing your pipeline - scanning the code, protecting secrets, securely managing code deployments, or even authentication and authorization mechanisms, but we rarely talk about all of these together.After years of being in the trenches and realizing that the attack surface is growing and the threat landscape becoming more and more complex, it has become increasingly apparent that security teams need to adapt and modify strategies to keep up with the new reality of CI/CD protection, without compromising developer velocity.In this talk I would like to propose a new way of thinking about CI/CD security - that encompasses the three disciplines that comprise CI/CD security - security in the pipeline, of the pipeline, and around the pipeline. Partial coverage of any or all of these disciplines simply will not cut it with the continuously evolving risk landscape. Security engineers need to address each of these aspects in their entirety to provide the full scope of coverage that modern organizations need, and I will take a deep dive on the challenges each introduce, and the approaches and techniques for mitigating them based on adversarial sec research.
Sponsored Ignites: Innocom, DoIT and Grafana DevOpsDays Tel Aviv 2026-07-01 | 00:00:00 - Innocom00:05:05 - DoIT00:10:52 - GrafanaInnocom: MTTR Vertigo: The Power of Consolidation.2 AM. Service down. You're flipping between seven tools trying to figure out where to start -that's MTTR vertigo. This talk shows how consolidating your entire stack onto one platform removes the silos that turn 5-minute investigations into hour-long ones. You'll see what root cause analysis looks like when metrics, logs, traces, and ownership all live in one database, queryable in one language. Walk in skeptical. Leave with a clear answer to why silos cost more than every observability license combined.Or Plavnik, Site Manager, InnocomMichael Isayev, Software Engineer, InnocomDoIT: Leaner K8s with PSI metricsPSI metrics have graduated to beta in K8s v1.34. It may be a small feature but it can matter a lot for your application performance and efficiency. Let's understand how they can make your clusters and autoscaling leaner.Anton Weiss, Developer Advocate, DoITGrafana: AI Makes Observability Conversational. Grafana Makes It Correct.Everyone is exposing an MCP server onto their observability stack. That's the easy part. Wiring an LLM to your metrics, logs, and traces takes an afternoon. The hard part is making the answers true, so that when you ask "why is service slow at 3am," you get a real root cause instead of a confident hallucination built on whatever the model happened to retrieve. This talk is about what separates a demo from something you'd trust during an incident: a deterministic foundation underneath the AI. Grafana pairs generative AI with a Knowledge Graph that holds ground truth, the real relationships between your services, pods, traces, logs, and the teams that own them. The AI doesn't guess at your topology. It reasons over a verified map of it. Same idea on the data layer: open standards, real telemetry, no black box.Danny Parizada, Solution Engineer, Grafana Labs
I Am the Law: How AI made rule-writing the most important skill in engineering, Daniel Avital DevOpsDays Tel Aviv 2026-07-01 | AWS TrackNobody ever cared about writing rules well. You'd drop some standards in a Confluence doc, maybe an onboarding wiki, and move on. It didn't really matter how precise they were because nobody was reading them anyway. As long as you had strong engineers and a decent review culture, the codebase held together well enough.AI changed the math. When code volume explodes and the author is increasingly a model, every other propagation mechanism breaks down. What's left is the rule. How precisely you can define it, scope it, and enforce it is now the difference between a codebase that reflects your judgment and one that slowly stops being yours.This talk is about that shift.
Raising the Signal: Predicting Performance Degradation with Automated RCA on AWS, Assaf Peleg DevOpsDays Tel Aviv 2026-07-01 | AWS TrackSOC & NOC teams aren't short on data - they're short on signal. With overwhelming alerts & logs, the challenge is improving Signal-to-Noise Ratio (SNR): isolating what truly matters.This session presents an automated RCA framework running on Amazon EKS that amplifies signal from network and security incidents. By combining trend detection with GenAI on Amazon Bedrock, the framework transforms complex network and security data into clear, actionable explanations of why incidents occur—and what is likely to happen next.Our customizable, tool-driven architecture builds tailored RCA workflows per story type, increasing SNR precision. Amazon Bedrock AgentCore provides secure runtime, contextual memory, observability, operationalizing high-SNR insights into RCA-driven actions.Detection creates data. RCA creates signal.
AI-Driven DevOps: From Operational Overload to Continuous Optimization, Asia Salner DevOpsDays Tel Aviv 2026-07-01 | AWS TrackWe operate a multi-account AWS environment (12 accounts, 4 regions) on EKS, Terraform, GitLab CI, and ArgoCD. The primary constraint was not infrastructure scale, but operational overload: growing backlog, missed cost optimizations, delayed risk detection, and constant reactive work.This session presents how AI is embedded into both DevOps workflows and the platform layer to shift from reactive operations to continuous optimization,--Cost & hygiene — continuous scanning across AWS accounts to detect unused resources, redundant services, orphaned assets, and cost inefficiencies, with fixes generated as code--Security & compliance — identifying over-permissioned access, configuration drift, and policy violations, with controlled remediation via pipelines--Production signal analysis — logs, metrics, and code analyzed to surface hidden issues such as memory leaks and abnormal patterns missed during development--Shift-left review — AI-assisted code review for Terraform, Kubernetes, and CI/CD to prevent misconfigurations before deployment--Autonomous task execution — DevOps tasks from Jira translated into reviewed, GitOps-driven changes--Impact-based prioritization — operational work ranked by cost, reliability, and security impact, reducing backlog and eliminating low-value manual triageAll actions are enforced through GitOps and review. No direct production mutations.Focus: architecture, integration patterns, control mechanisms, and measurable impact on cost, MTTR, and operational load.
Your AI Agent Is a Monolith — Heres How We Made Ours Modular, Haim Raitsev DevOpsDays Tel Aviv 2026-07-01 | AWS TrackMost AI agents start the same way: a single prompt, a growing list of tools, and increasingly fragile behavior as complexity creeps in. Ours did too - until it became unmaintainable. In this talk, I'll show how we refactored our production IT helpdesk agent into a composable toolkit architecture using AWS Bedrock, where each capability - password resets, app provisioning, device recovery, knowledge base search, ticket management - is an independent, self-contained module with its own tools, prompts, dependencies, and runtime activation gates. Skills assemble dynamically per-request based on tenant configuration, user roles, and identity provider capabilities. I'll cover the engineering patterns that made it work: runtime gating, lazy dependency injection, streaming responses, and how we keep a multi-tenant agent from doing things it shouldn't. Real production code, real lessons learned.
Running 100M prompts on EC2 under GPU capacity constraints, Sharon Dahan DevOpsDays Tel Aviv 2026-07-01 | AWS TrackThis session is a technical customer story about running a 100M-image multimodal inference workload on Amazon EKS with vLLM under real GPU capacity constraints. It starts with a straightforward capacity plan based on B200 machines and a GPU-hour calculation, then shows why that plan breaks down in practice when ideal capacity is not consistently available. From there, the talk explains how the hardware strategy expands across multiple AWS GPU families, why NVFP4 changes the economics, and why vLLM serving strategy has to be adapted to the topology of each machine type. The session closes with the final architecture and an additional slide on how Impala productizes this approach for production workloads.
Agentic Chaos - What 86K+ Agent Codebases Reveal About 700K+ Exposed AI Systems, Bar Kaduri DevOpsDays Tel Aviv 2026-07-01 | COMMUNITY TRACKOver the past two years, AI agents have been stealthily becoming the new backbone of the global internet infrastructure. Autonomous systems capable of invoking tools, executing code, orchestrating workflows, and interacting with external services are now being built and deployed across production environments, developer workflows and tools, automation platforms, data pipelines, and enterprise systems as a whole.What's become glaringly obvious is that despite the speed of this adoption, almost nothing is known about how these systems are actually built or secured in the wild.To answer that question, we conducted one of the largest first-of-its-kind empirical studies of the agent ecosystem. We analyzed more than 86,000 public repositories implementing agent logic across frameworks including LangChain, LangGraph, CrewAI, AutoGen, and Model Context Protocol (MCP). We examined prompt construction patterns, tool implementations, authentication models, execution capabilities, and permission boundaries to understand how developers are building agents in practice.We paired this code-level analysis with internet-wide infrastructure measurement using Shodan, Censys, and ShadowServer to map where agent platforms are actually running in the wild. The research surfaced more than 700,000 exposed agent-related systems on the public internet, including Ollama inference servers, Ray clusters, n8n automation platforms, and MCP tool servers. Many of these systems were directly exposed to the internet with little or no authentication and, in numerous cases, were vulnerable to known high-impact CVEs.Together, these two datasets reveal a striking pattern. The same architectural assumptions and security shortcuts visible in agent codebases appear repeatedly in real deployments at internet scale.This talk presents the data, visualizations, and insights produced from this research.Using examples drawn directly from the dataset, we reconstruct several representative attack paths created by common agent design patterns. We show how seemingly harmless implementation choices, such as tool exposure, prompt construction shortcuts, and weak capability boundaries - can cascade into exploitable conditions once agents are deployed in real environments. We'll wrap up with practical architectural changes that agent frameworks and platform teams can adopt to prevent these patterns from becoming the next generation of supply-chain vulnerabilities.
Why Your Kubernetes Cluster Will Fail: Lessons from 1 Million Real-World Incidents, Asaf Savich DevOpsDays Tel Aviv 2026-07-01 | COMMUNITY TRACKWe've analyzed over 1 million production K8s failures across thousands of clusters. The data reveals something striking - the vast majority of incidents fall into predictable, preventable categories. By the law of large numbers, if we address these recurring issues, we can drastically improve production reliability.This talk presents the most common K8s failure patterns backed by real data at scale. We'll cover 6 major categories - Resource Exhaustion (OOMKilled pods, memory leaks, GPU thermal throttling), Image & Deployment Issues (ImagePullBackOff, stuck rollouts), Config & Secret Management (rotation breaking apps, ConfigMap drift), Cascading Failures (missing ConfigMaps triggering multi-pod failures, dependency chains), Storage & Persistence (PVC conflicts, CSI driver issues), & App vs Infra Debugging (CrashLoopBackOff mysteries, GPU XID errors).For each category, we'll show real K8s events, explain why these failures are so common, and provide actionable prevention strategies.
The Agentic Era: Microservices, Reimagined?, Issac Goldstand DevOpsDays Tel Aviv 2026-07-01 | COMMUNITY TRACKAs we enter the 'Agentic Era' of software engineering, we are replacing deterministic programming languages with natural language and replacing traditional flow control with non-deterministic LLMs. But are we throwing away a decade of architectural best practices? While the talk will focus on high-level architecture principles, we will briefly dip into code written using the AWS Strands SDK to understand the basics of writing AI Agents that can run on AWS Bedrock. We'll also understand how this relates to writing code with Agentic AI IDEs, such as AWS Kiro or VS Code.
From Black Box to Glass Box: Observability for LLMs, Aviv Zohari DevOpsDays Tel Aviv 2026-07-01 | AWS TrackIn just a few years, LLMs have gone from research curiosities to the backbone of new software experiences. Organizations are rapidly productionizing LLM workflows because of their immense value, but often without observability guardrails. This introduces new layers of fragility and complexity: performance volatility, quality drift, and security risks. In this talk, we explore how to monitor and troubleshoot LLM applications with zero instrumentation. Whether on commercial LLM stacks or AWS Bedrock, we'll break open the LLM black box and learn how to track token usage, response latency, data exposure risks, and model execution failures.
What Production Knows: Closing the Loop Between AI Agents and the Systems They Build, May Walter DevOpsDays Tel Aviv 2026-07-01 | COMMUNITY TRACKMost of the AI-for-engineering conversation runs in one direction: humans tell agents what to do, and agents produce code. But for high-velocity teams, the more interesting flow runs the other way, production telling the agents what is actually happening, so they can fix, refactor, and keep shipping without drifting away from reality.This talk is about that reverse loop. It looks at what changes when the systems we operate start moving faster than the people who own them, the failure modes that quietly become normal when agents work without ground truth, and the durable plumbing that keeps them honest: traces, error budgets, customer signal, SLO breaches, incident timelines, all fed back as first-class inputs to the tools doing the writing.If the broader AI shift is changing what it means to author code, this is the operational half of that story: the feedback infrastructure that makes the new way of working actually safe to run.
SPOTLIGHT TALK: Whats happening to Open Source?, Itay Shakury DevOpsDays Tel Aviv 2026-07-01 | Open Source has always been a delicate ecosystem, a rare coexistence of interests, passion and opportunity. That delicate balance is being tested now, and the first signs are starting to show. You might have felt it, I know I have, a gradual change in my own experience as user and maintainer. Small things that only when combined reveal a larger pattern. I can now see four clear trends that I believe are happening right now, and are fundamentally changing the open source landscape. In this talk, I will share my observations and insights about what and how the open source experience is changing, and also why it's happening now (spoiler alert: AI). You will learn to navigate the new open source landscape, the new practices and culture that is emerging, the tools and policies that are in use, and the way it affects you, as an open source user, contributor, or maintainer.
KEYNOTE: Your Agent Built It. Who the 🤬 Runs It?, Ran Tavory DevOpsDays Tel Aviv 2026-07-01 | Coding agents are shipping production code faster than you can read it. Somebody still has to run it at 2 AM. This talk is for the people on that pager - three forces making production harder, and the concrete shifts (observability, SLOs, policy-as-code, AI investigation reports) that define the next decade of the SRE role.
RAG on Your Data Lake, Haggai Zohar DevOpsDays Tel Aviv 2026-07-01 | COMMUNITY TRACKMost data lakes aren't LLM-ready. Without structure, metadata, and governance, connecting generative AI to S3 data creates risk, cost, and unreliable results. This session walks through an end-to-end Retrieval-Augmented Generation (RAG) architecture on AWS. We'll cover ingestion and preprocessing of structured and unstructured data from Amazon S3, building embeddings with Amazon Bedrock or SageMaker, and choosing vector storage with OpenSearch or Aurora pgvector. You'll see how retrieval orchestration and inference run securely with Bedrock, alongside IAM, PII, and governance controls. We'll also examine cost, latency, freshness trade-offs—and when RAG beats fine-tuning or search-only approaches.
Ignite: Thousands of Clusters Maintained: Lessons Learned and Tales to Tell, Itamar Syn-Herskho DevOpsDays Tel Aviv 2026-07-01 | In this talk, we'll dive deep into the heart of OpenSearch and higlight common mistakes but also not so common gotchas that'd help you optimize and stabilize your deployment too. This talk is based on 15 years of cluster maintenance in production - at every vertical, any scale, all use-cases you can think of. From Elasticsearch 0.10 to OpenSearch 3.0, I've seen it all and survived to tell the story.
The Last Human On-Call — When AI fixes everything, who still knows anything?, Miki Manor DevOpsDays Tel Aviv 2026-07-01 | Eliminating toil is one of the SRE community's most sacred goals. AI agents are finally delivering on that promise — handling first-line incident response, running runbooks, auto-remediating known failure modes. Page volume is down. Mean time to resolution is down. On-call burden is down. This is unambiguously good. And it's also hiding a slow-moving crisis. SREs have always learned production systems by living with them — by being paged at 2am, clicking through runbook steps that felt tedious, noticing that this particular service always spikes after a deploy to that downstream dependency. That pattern recognition didn't live in documentation. It lived in the repetition. It lived in the toil.When AI agents absorb all of that, the humans who remain are faster, less burned out, and increasingly de-calibrated from their own systems. The first time something genuinely novel breaks — outside the agent's training distribution, no runbook applicable — you need a human who deeply knows the system. That human may no longer exist because they haven't been paged in eighteen months. This talk is about the expertise debt accumulating silently in every org embracing agentic operations — and the engineering practices that can help you not wake up to it the hard way.
We Tried OSS Model Routing Tools So You Dont Have To, Niv Yungelson & Hai Rozencwajg DevOpsDays Tel Aviv 2026-07-01 | COMMUNITY TRACKModern AI applications are no longer single prompts. They are pipelines of decisions: retrieval, reasoning, tool use, and execution. However, most systems still default to a single model across all steps, leading to unnecessary cost, latency, and failure modes.This talk presents a hands-on evaluation of open-source model routing tools in the context of multi-step AI workflows. Instead of comparing models in isolation, we examine how routing decisions change when facing a multi-step pipeline as part of computer-use task execution.We'll cover real experiments, architectural patterns, and common pitfalls, and provide a practical approach to step-aware routing that improves both performance and cost efficiency.
AI Talks IL #1 with Boaz Ziniman and Chen Feldman, hosted by Grove Ventures DevOpsDays Tel Aviv 2026-06-30 | For our very first event we welcomed to our stage two fantastic speakers, who helped get the teams feeling a combination of both FOMO for how to get started, and FUD for adopting AI right - up to speed on the terminology and team adoption strategies.00:00:00 - Welcome and TLVCommunity Events - Sharone Zitzman00:02:15 - Opening Words by Lotan Levkowitz, Grove Ventures00:05:55 - It's All One Loop, Boaz Ziniman00:37:00 - Spark, Embed, Own, Chen Feldman It's All One Loop, Boaz Ziniman Boaz deystifies the difference between all of the tools that seem widely adopted by everyone in the ecosystem, and you're still trying to figure out - from the Skills, to the Agents, MCP and hooks.Spark, Embed, Own, Chen Feldman Chen shares what he's learning embedded with engineering orgs right now - the gap between scattered individual AI usage and team-owned practice, what actually compounds across a team, and the metrics that genuinely predict whether adoption sticks.
Truly Cloud Native AI Agents with Kagent and Khook, Anton Weiss DevOpsDays Tel Aviv 2025-12-29 | AI agents are more than just LLMs with tools.In this talk I'll discuss the cost of AI agency - all we need to do to make AI agents useful and secure.I will then present Kagent - the CNCF project that makes agents cloud native and Khook - the reactive component for Kagent I've contributed and am now maintaining. The listeners will walk out with practical tools for building and running reactive AI agents on K8s.
Behind the Cloud @Twitter 1.0, Bobby Dorlus DevOpsDays Tel Aviv 2025-12-29 | As an SRE, capacity management, including demand forecasting, is a challenging routine when managing one of the world's largest Aurora/Mesos compute clusters. While Twitter Engineering moves forward with re-platforming to Kubernetes. Come hear me speak about implementing a technical solution that increased our existing clusters capacity by 9%—resulting in a multi-year saving of 9-figures in CAPEX by solving inefficiency challenges at data center scale.● Discovering underutilized resources (CPU, Mem, and Network) within our storage platforms.● Implementing a modern solution based on past experience with: Solaris Zones, Cgroups and Namespaces● Colocate compute and storage workload on the same bare-metal machine
The Anatomy of a Patch: Backporting CVEs Without Breaking Things, Benji Kalman DevOpsDays Tel Aviv 2025-12-29 | When a new CVE drops, applying the fix isn't as simple as merging upstream changes––they hit live systems with layers of legacy, dependencies & constraints.Welcome to the shadowy underbelly of backports, a delicate operation where the goal is to surgically apply a fix, without disrupting the stability of older releases, prod systems or their intricate web of dependencies.This talk dissects CVE-2024-37370 in MIT Kerberos - a vuln that can be found in all of the major Linux variants and underpins cloud-native applications broadly, including many CNCF projects.We'll explore what constitutes a CVE-worthy vulnerability, what the vulnerable code actually looks like, and will reverse engineer how it's fixed. Then we'll walk you through step by step how to reverse-engineer a backport: unpacking the challenges of retrofitting fixes into older codebases, rewriting patches to accommodate shifting APIs, and avoiding side effects that ripple across documentation, test suites, & configs.
We need to talk about limits, Avishai Ish-Shalom DevOpsDays Tel Aviv 2025-12-29 | For years we have been bound by annoying restrictions, but the age of cloud ushered a limitless world where almost anything seems possible. Unfortunately, nothing in life is that easy and we discovered the hard way that although our tech credit cards may not have a spending limit, that doesn't mean there isn't a pay day...We have become spoiled brats, forgetting the warning of our parents: "you gotta have limits!". This talk is about the long lost engineering principles of resource management which have become oh so relevant in these troubled times.
The Hidden Complexity of Time in Serverless: A 5-Minute Reality Check DevOpsDays Tel Aviv 2025-12-29 | Picture this: A government announces a last-minute timezone change with just days of notice. Suddenly, your serverless application has the wrong time assumptions baked into immutable functions. Users get notifications at wrong times, business logic breaks, and support is flooded with "time bug" reports.This happens 3-6 times per year as governments adjust timezone rules for political, religious, or economic reasons. In traditional servers, you'd update the timezone database and restart. In serverless? Your timezone data is frozen into deployment artifacts.This lightning talk shows the one hybrid architecture pattern that actually works in production - preventing millions in downtime when governments decide to mess with time.
Adjusting your mirrors: Finding and fixing blind spots in your configurations, Dor Meiri DevOpsDays Tel Aviv 2025-12-29 | Most systems rely on default configuration values, and that’s usually fine. But these defaults often go untouched for years, even when they no longer match the system’s scale or behavior. The result? Silent performance losses that are easy to miss and hard to trace.In this talk, I’ll share the true story of how changing a single digit in an obscure timeout setting cut our API latency in half. The issue wasn’t bad code, it was an outdated assumption baked into a default value that didn’t fit our system anymore.You’ll learn:- How configuration defaults can quietly create performance bottlenecks- What invisible inefficiencies look like in an advanced DB connections management (e.g. sharding, connection pools, connections multiplexer)- How to detect these issues in production systems, even when logs and metrics seem fine- A practical approach for evaluating, tuning, and monitoring configuration selectionThis talk is for anyone working at the intersection of development and operations. If your systems are “working” but not performing, this will help you uncover why, and what to do next.
Happy Hour AMA with Solomon Hykes and Bobby Dorlus - Moderator: Gil Zellner DevOpsDays Tel Aviv 2025-12-29 | Now is your chance to tap into the builders from our community - Solomon Hykes, creator of Docker and Bobby Dorlus, one of the first SREs at Twitter - an open, mic and AMA - to find out all the juiciest details you can only ask in person.
Build a Self-Service Hub in Slack, Shaked Braimok Yosef DevOpsDays Tel Aviv 2025-12-29 | In this quick hands-on session, we’ll explore how to build and deploy a custom self-service hub directly in Slack, empowering your team to handle day-to-day operational tasks without needing to rely on DevOps or platform engineers. Whether you're automating routine infrastructure tasks, managing secrets, or facilitating service requests, Slack can become your team's one-stop-shop for all things self-service.You'll walk away with the skills to:Design and implement Slack bots that integrate with your internal toolsSet up Slack commands to automate common workflows and improve team productivityLeverage Slack APIs and popular frameworks like Slack Bolt to streamline day-to-day tasksEmpower your team to take control of their operational processes while reducing bottlenecks and dependenciesThis session is ideal for DevOps professionals, platform engineers, and anyone looking to enhance team autonomy and efficiency through automation and self-service tools in Slack.
5 Serverless Patterns You Should Stop Using (And What to Do Instead), Ran Isenberg DevOpsDays Tel Aviv 2025-12-29 | We’ve all used them—architectural patterns that seem like the right choice for Serverless architectures. I did, too. At first, they felt efficient, straightforward, and even best practice. However, as my Serverless services matured, complexity kicked in, and I started to see the downsides: increased complexity, maintenance headaches, and even security and performance challenges.In this session, I’ll walk you through five Serverless patterns I once used (and regretted): API Gateway direct integrations, monolith Lambdas, direct Lambda-to-Lambda invocations, writing all code in the handler, and using the wrong EventBridge service for scheduling.You’ll learn why these patterns can be problematic, when they make sense, and what to use instead.
Supercharging Platform Engineering at Scale: The Dell Automation Platform, Moshe Mizrachi, Dell DevOpsDays Tel Aviv 2025-12-29 | Join us for a lightning tour into how automation frameworks and AI-driven orchestration are reshaping infrastructure. The Dell Automation Platform is a foundational enabler for platform engineering, empowering your team to deliver any service (XaaS) or IT environment, whether in data centers or the cloud, with the agility and simplicity of a public cloud experience. In just ten minutes, you’ll get a glimpse of tools that make innovation unstoppable.
A Perfect Cluster, A Kubernetes Optimization Framework w/ PerfectScale by DoiT A. Weiss & A. Berezin DevOpsDays Tel Aviv 2025-12-29 | Once your Kubernetes environment is live, the real work begins. This is Day-2 Operations—where stability, performance, and cost efficiency must be maintained continuously. Optimization at this stage isn’t about one-off fixes. It’s about creating a repeatable and proactive process.In this short, practical demo session we’ll present a 6-step framework we’ve developed to help our customers cut waste, improve reliability, and keep their Kubernetes clusters lean and responsive.
Thinking Outside the Compositions: When Control Planes Logic Becomes the Bottleneck, Elhay Efrat DevOpsDays Tel Aviv 2025-12-29 | Crossplane promises a unified API for cloud infrastructure, but many teams hit a wall as they scale. Your Compositions, initially simple, swell with complex patches and functions, creating a dangerous trade-off; you achieve a simple developer interface by hiding complexity in a system that becomes opaque, slow, and nearly impossible to debug. So, how do you trace a production failure through layers of runtime "magic"?In this session, we'll present cell architectural patterns that resolves this conflict by shifting dynamic logic out of runtime Compositions and into a dedicated, build-time pre-processor - easy dev, test, validate. We'll show you how to decouple the developer's high-level intent from the final, low-level infrastructure specification, giving you the best of both worlds: a great developer experience without sacrificing platform visibility.
How to Build Quality-Driven Agentic AI in Noisy Big Data Environments, Itiel Shwartz DevOpsDays Tel Aviv 2025-12-29 | Building reliable agentic AI systems in prod environments presents unique challenges when dealing with massive, noisy datasets. This talk shares hard-won lessons from developing an AI agent that processes millions of K8s events daily to deliver autonomous troubleshooting that reached 95%+ accuracy in benchmarking.The fundamental challenge isn't LLM capability—it's building systems that maintain reliability when 90% of your data is noise. We'll explore why most agentic AI fails in production: hallucinations masquerading as insights, inability to validate reasoning chains, and the brittle nature of RAG systems when dealing with complex, interconnected failure modes.Gain practical tips learned through painful prod iterations: how to build validation frameworks that catch LLM errors before they reach users, architectural patterns for constraining problem spaces without losing effectiveness & methods for creating evidence-based reasoning that can be audited & improved systematically.
Dancing with Failure - The Art of Timeouts & Retries, Alon Nativ DevOpsDays Tel Aviv 2025-12-29 | “Timeouts, Retries, and the Devil” dives into the often-overlooked complexity of error handling in high-scale systems. We’ll explore why short timeouts are essential for resilience—but dangerously easy to misuse—and how naive retries can spiral into system-wide meltdowns. Using real-world examples, including a postmortem-style analysis of a production incident inspired by real world problems, I’ll walk through actionable strategies to design smarter retries, prevent service hammering, and avoid cascading failures. This is a talk full of practical wisdom, gotchas, and system design insights that will stick with you the next time you’re writing retry().
All the Vibes, Zero Anxiety - Safe AI Coding on Real DBs, Amir More, Baseshift DevOpsDays Tel Aviv 2025-12-29 | AI assisted coding moves fast, sometimes too fast. One stray suggestion can drop a table or break a migration.Baseshift removes that anxiety by giving developers safe, production-like database copies that are anonymized, subsetted, and instantly resettable, with full visibility into every change. So you can vibe-code fearlessly, recover instantly, and ship with confidence.
Agentic AI as an Accelerator for Infrastructure Operations, Jonathan Viduchinsky, Quali DevOpsDays Tel Aviv 2025-12-29 | At their best, tools should feel like teammates — and AI for infrastructure is no different. In this session, we’ll show how Quali’s Torque MCP Server uses rich environment context to let agents safely provision and clean up infrastructure with traceable, governed actions. Through a live demo, you’ll see how environment-aware agents turn repetitive infra work into fast, reliable, and auditable automation
Disaster Recovery in the Serverless Realm, Orel Bello DevOpsDays Tel Aviv 2025-12-29 | What happens when a serverless application faces a regional outage? Traditional DR strategies don’t always translate to a cloud-native environment, requiring a different approach. This talk explores four key DR strategies—Backup and Restore, Pilot Light, Warm Standby, and Multi-Site—and their trade-offs in cost, complexity, and recovery time. Using real-world scenarios, we’ll discuss multi-region failover, automated recovery with infrastructure-as-code, and best practices for resilience in serverless architectures. Attendees will leave with a practical framework for designing scalable and cost-effective DR strategies in a cloud-native world.
Robusta.dev: HolmesGPT - the CNCF AI Agent for SREs, Natan Yellin - Robusta.dev DevOpsDays Tel Aviv 2025-12-29 | Just how good are AI agents at troubleshooting cloud alerts and preventing downtime? In this demo, we’ll throw HolmesGPT - the new CNCF AI Agent - at complex real world problems and try to break it. We’ll stress test on a wide range of scenarios - from Kubernetes problems, to AWS issues, to application bugs - and we’ll compare model providers like OpenAI, Anthropic, and DeepSeek. Come join us for some fun as we take the latest CNCF project for a test run!
Building a Production-Grade AI/ML Inference Platform on Kubernetes, Liad Drori DevOpsDays Tel Aviv 2025-12-29 | Building an AI/ML platform that can analyze and summarize massive volumes of unstructured medical data in real time requires more than just powerful models, it demands a production-grade infrastructure capable of handling complex inference workloads at scale.This session dives into how to design and operate AI inference workloads on Kubernetes through an Amazon EKS example, for real-world production environments. It will cover model handoff processes from data science teams, validation and readiness assessments, deployment architecture, scheduling and scaling strategies, and inference traffic management. The session will also explore observability, performance tuning, and optimization techniques that are essential for maintaining reliability under heavy load.Walk away with a deep understanding of the operational challenges, trade-offs, and engineering solutions involved in running high-performance AI/ML inference systems efficiently and responsibly at scale.
Fireside Chat: Unpacking the Mega Cyber-Success - The Story Behind the Tech, Melio, Wiz, CyberArk DevOpsDays Tel Aviv 2025-12-29 | Or CohenVP ArchitectureMelioDaniel SchwartzerChief Product TechnologistCyberArkOron NoahVP ProductWizOr Cohen, VP Architecture at Melio will be joined by Daniel Schwartzer a Sr. Director of Data and Horizontal AI at CyberArk, and Oron Noah, VP Product at Wiz to talk about their incredible journeys to billion dollar exit success - and the technology out of our very own community that made it possible.
Beyond Argo Events: Leveraging NATS for Scalable Webhook Management in Dynamic Environments,Or Navon DevOpsDays Tel Aviv 2025-12-29 | Managing webhooks at scale is a complex challenge, especially in dynamic environments. Supporting dozens of dynamic development environments introduced bottlenecks due to integration limitations and reliability concerns.Existing tools, including Argo Events, fell short of meeting our requirements for persistence, observability, and simplicity.In this talk, we will explore how we used NATS, a lightweight and scalable message queue, to build a custom solution for webhook management.We'll cover how you can achieve exactly once delivery, fault tolerance, and resource efficiency (40 MB memory, 0.1 CPU) while drastically improving webhook routing and testing across environments.You will gain practical insights into using NATS for event-driven systems, evaluating CNCF tools, and building tailored solutions that reduce complexity while enhancing scalability.
Keynote: Building Docker: Behind the Scenes of the Container Revolution, Solomon Hykes DevOpsDays Tel Aviv 2025-12-29 | Tap into the mind of the person who built one of the most useful tools in the DevOps ecosystem - Docker - that fundamentally revolutionized the way we build and deliver software. He's joining our community to geek out on all the technical thoughts that led him to build Docker (some that he has never shared publicly before), the visionary roadmap vs. what we use today, and how all of that led him to his latest project Dagger - where he's trying to fix everything that's still broken and painful in CI. Expect it to be nerdy, filled with code and lore, and come away understanding how to think about codifying the rigors of systems engineering into our day to day work, stack, and tools.
Moving Mountains, Mike Fiedler DevOpsDays Tel Aviv 2025-12-29 | Software lifecycles extend far beyond the initial build phase. Often a system built in a short time may continue to serve for years longer. These become mountains, foundations for a business or product, and contains encoded decisions, bug fixes, and implicit behaviors that can prove difficult to maintain.Migrating large systems can be risky and daunting, and the strategy to migrate a system changes significantly based on the requirements.In this session, Mike will share the approach taken over two years to migrate 6 core systems into a single source of truth, sharing techniques, challenges, and resolutions that worked for overall success.
Tracing the Untraceable: Full-Stack Observability for LLMs and Agents, Shachar Azulay, groundcover DevOpsDays Tel Aviv 2025-12-29 | In just a few years, LLMs have gone from research curiosities to the backbone of new software experiences. Organizations are rushing to productionize LLM workflows because of their immense value — but they’re doing so without observability guardrails. This introduces new layers of fragility and complexity: performance volatility, quality drift, and security risks.In this talk, we’ll explore how to instantly monitor and troubleshoot LLM applications with zero instrumentation. Whether on commercial LLM stacks or AWS Bedrock, we’ll break out of the LLM black box and learn how to monitor token usage, response latency, data exposure prompts, and model execution failures.
Newsflash: There is no Quality as a Service, Niv Yungelson DevOpsDays Tel Aviv 2025-12-29 | Our hardware is getting faster, why the heck does our software get slower?our servers are virtual at best. Our databases use storage that is not managed by us and our code is on someone else’s domain.In this talk we will speak the unspeakable- over-abstractions can hurt the quality of our product and we don’t do anything about it.I will give examples of before and after, and alongside to a lot of food for thought, we will give some practical tips on how to be quality minded in a world of SaaS
From Duct Tape to Declarative:Playtikas Platform Overhaul at Unicorn Scale,S.Rosenberg & S.Mashiach DevOpsDays Tel Aviv 2025-12-29 | Scott RosenbergLead ArchitectTeraskyShemer MashiachDirector, IT Platforms Solutions / ArchitectPlaytikaSuccess is a double-edged sword. As the gaming industry boomed, Playtika’s platform couldn't keep pace. What was once manageable became chaos: Ansible scripts, imperative Jenkins pipelines, a sprawl of internal portals & everyone doing automation their own way.Environments took days to spin up.Onboarding was guesswork.DON'T touch the fragile parts.This talk unpacks how Playtika modernized that mess into a CNCF-driven platform stack. Learn how they moved to declarative K8s clusters with ClusterAPI, GitOps deployments with FluxCD, and infra managed via Crossplane. Internal tools were replaced with a single Backstage-powered portal, delivering developer efficiency at scale.But the real shift wasn’t just technical—it was cultural.Ticket ops became self-service. Teams aligned. Provisioning dropped from days to under an hour all with OSS.If your platform feels like it’s held together with duct tape and hope, this talk shows how standardization isn’t boring—it’s what scales.
Oops-Driven Development, Shahar Shporer DevOpsDays Tel Aviv 2025-12-29 | DevOps is all about automation, speed, and reliability - until your pipeline skips a step, your Helm chart misbehaves, or your IaC plan decides to delete production.Oops.In this bite-sized lightning talk, we’ll laugh (nervously) at the kinds of mistakes that happen when we don’t test our infrastructure, pipelines, and configs- and then we’ll look at how to stop them. We’ll cover best practices and real tools for testing your DevOps work and share ways to shift testing down into the layers most people forget.Because let’s be honest: if your job is to keep everything running, so you probably shouldn’t be deploying blindfolded.
Is your database holding you back? Challenge your perception of NoSQL performance!, Aerospike DevOpsDays Tel Aviv 2025-12-29 | Oshrat Ben-Avi ZabdulovitzZohar ElkayamJoin us for a high-impact, live demonstration where we showcase Aerospike taking on a massive, real-world mixed read-write workload and handling it without breaking a sweat. We’ll begin by revealing a compelling full user profile use case with 1.3 billion records and 6 TB of replicated data - a scale that challenges traditional databases. Watch what happens next: our cluster effortlessly manages this massive scenario, sustaining an incredible 400k–800k operations per second while delivering predictable, rock-solid latencies. How solid? 99% of requests are served in under 1 millisecond! This efficiency is achieved by using only 2.5% of the total data size for memory.This demo is proof of how you can achieve extreme scale, maximize efficiency, and unlock massive cost savings on standard cloud infrastructure. Your data strategy starts here - don't miss this opportunity to see true power.
Orchestrating Autonomous Agents in DevOps: Comparing Strands Agents Deep Agents & AutoGen,Engin Diri DevOpsDays Tel Aviv 2025-12-29 | AI agents are moving from experiments to real tools in DevOps. But with so many frameworks to choose from, it can be hard to know which one fits your needs.In this talk, we look at three leading multi-agent AI frameworks: Strands Agents, Deep Agents, and AutoGen. We will show how each handles Model Context Protocol (MCP) integration and orchestration for autonomous workflows.You will see:- Strands Agents – model-driven orchestration, provider-agnostic setup, built-in MCP support, and agent-to-agent handoffs.- Deep Agents – graph-based workflows, agent swarms, memory management, and scaling to production.- AutoGen – event-driven collaboration, conversational agents, and the rapid prototyping power of AutoGen Studio.We will walk through real DevOps use cases like CI/CD automation, incident response, and infrastructure provisioning.By the end, you will have a clear view of the strengths and trade-offs of each framework, along with a simple decision guide to choose the right one for your team.
Would You Drive a car Without a Dashboard?, Gabriela Nir DevOpsDays Tel Aviv 2025-12-29 | You wouldn’t drive without a dashboard. So why do so many engineering teams run without any real operational visibility?As Engineering Leaders, we have so much on our mind, and we need a tool, like in the car, that will allow us to drive to our destination, with a dashboard focusing us on what matters.In this talk, I introduce a simple, 3-layer model for building dashboards that are actually helpful:1. High-Level View – Are we on track?2. Detailed View – Where’s the work?3. Red Lights Focused– What’s broken?
Launch Partys Over.Now What? A Guide to Real-World Ops w/ Crossplane & ACK, Guy Menahem & Adam Kobi DevOpsDays Tel Aviv 2025-12-29 | The future of platform engineering is a single, unified control plane, where Kubernetes-native tools like Crossplane and ACK manage both applications and infrastructure through one consistent GitOps workflow. However, the path from this powerful vision to a stable, production-ready reality presents a new class of "Day 2" operational challenges. After the initial launch party, teams face a fundamentally different operational model where the familiar safety nets of traditional IaC no longer exist, and the rules for managing risk have completely changed.A common engineering incident begins not with a bang, but with a small commit—a typo that a traditional, Terraform plan would have caught instantly, but in the new "plan-less" world, it gets silently reconciled, leading to cascading failures. This session moves beyond the "hello world" to dissect these widespread failure patterns, exploring the architectural anti-patterns that emerge and presenting a playbook of field-tested solutions for operating declarative, continuously-reconciling systems at scale.This is not a theoretical guide; it is a tour of hard-won lessons from the community. Attendees will leave with a playbook for solving the most critical "Day 2" challenges, learning: a GitOps-native pre-flight check pattern; how to build a 'developer-portal dashboards for allowing developers to use it safely and a step-by-step strategy for safely migrating brownfield resources without downtime.
