Andrey Devyatkin’s Post

AI agents can take action. That changes everything - and most teams aren't ready for it. That's why we started Humans in the Loop - a video series for engineers and DevOps teams figuring out the agentic AI era without losing control of their systems. The first episode is out. Andrey Devyatkin and Fernando Gonçalves set the stage: what agentic AI actually means, why context is everything in infrastructure troubleshooting, and what tools like Cursor, MCP, Claude Code, and Amazon Q CLI mean for DevOps engineers today. https://lnkd.in/eiSbrP7Y

AI is showing up everywhere in DevOps—but rarely as a connected system. Code, testing, deployment, and ops are evolving with AI, but the real challenge is how they come together in production. On April 25, Harness and The AI Collective are bringing together practitioners shipping AI across the lifecycle and navigating real production complexity. Register to hear what actually works when AI moves into production: https://lnkd.in/gNAhE-3S

AI is showing up everywhere in DevOps—but rarely as a connected system. Code, testing, deployment, and ops are evolving with AI, but the real challenge is how they come together in production. On April 25, Harness and The AI Collective are bringing together practitioners shipping AI across the lifecycle and navigating real production complexity. Register to hear what actually works when AI moves into production: https://lnkd.in/gZZsJvKt

AI is showing up everywhere in DevOps—but rarely as a connected system. Code, testing, deployment, and ops are evolving with AI, but the real challenge is how they come together in production. On April 25, Harness and The AI Collective are bringing together practitioners shipping AI across the lifecycle and navigating real production complexity. Register to hear what actually works when AI moves into production: https://lnkd.in/gaYwbhJJ

We’re entering an era of code abundance. With AI generating more code than ever, the developer’s role is shifting from writing every line to curating, validating, and confidently shipping the right code. The result of this avalanche of code? Testing becomes the bottleneck. That's why we created CloudBees Smart Tests, which uses AI/ML to predict which tests matter most so teams can focus on risk, not volume. The outcome: faster feedback, leaner pipelines, and greater confidence in every release. Read more about Smart Tests on DevOps.com 🔗 https://lnkd.in/e9QqFgwf

Why Governance Determines Whether Agentic AI Accelerates or Stalls Engineering : AI coding tools offer rapid productivity gains, but "governance debt" often slows delivery. Learn how to embed risk-based controls and auditability into agentic AI workflows to scale engineering capacity. Read more: https://lnkd.in/gnFSPp3B 📚 Expand your DevOps knowledge! Join our community for continuous learning and skill development.

AI coding tools are undeniably speeding up development, but the rest of the delivery process is lagging behind. The 2026 State of DevOps Modernization Report from Harness compiles insights from 700 engineers, revealing that while AI is solving many challenges, it's also introducing new ones. 📄 Dive into the full report to explore AI's impact on DevOps: https://lnkd.in/eYu9PZNx

GitOps adoption is no longer the question. 🤔 Argo CD now powers a significant share of Kubernetes deployments, but as teams scale, the real challenge is no longer deploying code. It is promoting it across environments in a way that is consistent, visible, and reliable. 🐙 In this interview with TFiR, Hong Wang breaks down how that shift is playing out in real-world systems, and why continuous promotion is becoming a critical layer in modern delivery pipelines as AI accelerates development and increases deployment volume. 🔄 It is a clear look at where platform engineering is heading next. 👁️ Full interview in the comments. 👇 #GitOps #ArgoCD #Kargo #PlatformEngineering #CloudNative #AI

Day 14 of my DevOps → MLOps journey 🚀 Today I learned about Model Registry and Versioning, which is critical in production ML systems. Key idea: 👉 ML models need version control just like code. Using a model registry, we can: • track different model versions • manage lifecycle (staging → production) • rollback if something goes wrong Big insight: MLOps is not just about building models, but managing them safely in production. Next step: understanding CI/CD for ML systems. #MLOps #MachineLearning #MLflow #DevOpsToMLOps #AI

76% of DevOps teams added AI to their pipelines last year. The DORA report says some of them got worse, not better. Turns out AI code generation without automated testing is just shipping bugs faster. We wrote a DevOps tools guide that skips the 50-tool listicle format. Each tool is explained by pipeline stage, with picks by team size and a section on what AI tooling is actually ready for production. Come on, read it. https://lnkd.in/djg-kGv8