MLCON

What actually is DevOpsCon? Five days, five ways to learn. DevOpsCon Munich brings together software engineers, DevOps practitioners, and platform teams around the topics shaping modern software delivery: CI/CD, Kubernetes, platform engineering, and DevSecOps. The program is built so you can go as deep as you want: 🤖 AI Platform Engineering Day, building platforms for AI-native enterprises 🏗️ Platform Engineering Summit, scaling internal developer platforms 🎤 Main Conference + Expo, expert talks and keynotes 🛠️ Hands-on Workshops, practical deep dives 📚 Two-Day Bootcamps, intensive training From November 30 to December 4, in Munich. Explore the full program: devopscon.io/munich/program-… Explore AI Platform Engineering Day : devopscon.io/ai-platform-eng… #DevOpsCon #DevOpsConMunich #DevOps #PlatformEngineering #Kubernetes #CICD #DevSecOps #CloudNative

An AI agent is told to deploy a new microservice to production. It bypasses the portal entirely and goes straight for the platform’s APIs. Here’s the part that matters: it provisions the infrastructure, reads the system context to wire up the right database, and then gets blocked, waiting safely until a human approves the final push. That’s an agent-ready platform working as designed. Three things make it possible: ➡️Capabilities the agent can invoke. ➡️Context so it knows how your systems connect. ➡️Controls that stop it before it does damage. Our latest article breaks down how to build this into your platform. devopscon.io/blog/internal-d… #MLcon #AIAgents #PlatformEngineering #MLOps #AIEngineering #MCP

🤖 Can your governance keep up with autonomous AI agents? AI agents aren't just making recommendations anymore. They're accessing files, executing commands, and taking action across systems. But as agents gain more autonomy, one question becomes critical: Who's still in control? The new MLcon Magazine Vol. 24 explores how engineers can build secure, accountable, and governed AI systems. 🔐 Agentic AI Governance: Learn why delegated authority requires an Authorize–Monitor–Contain approach. 🛡️ Agents & MCP Security: Explore sandboxing, permission boundaries, command validation, and approval gates. ⚙️ Governed AI Workflows: Discover how preview-before-commit, validation, and human oversight help manage risk in FinTech and insurance. 💡 The key takeaway? Governance can't remain a policy on paper. It needs to be built into the systems where AI agents actually operate. 📖 Explore the new MLcon Magazine: devm.io/magazines/ml-magazin… 💬 Want to continue the conversation? Join us at MLcon Berlin, part of AI Native Week, to exchange ideas with fellow engineers and explore real-world approaches to responsible AI. 🔗 mlconference.ai/berlin/progr… #MLcon #AgenticAI #AIGovernance #AISecurity #MLOps

📌 Building private AI agents is easy. But who maintains them when every department deploys its own? 🎬 Insights from Tim Frey and Christian Schmitt, recorded live at MLcon. THE QUICK BREAKDOWN: 💡 Y Pipe aims to democratize local AI by connecting small models with business functions, enabling departments to build AI-powered processes using their own domain expertise. 💡 Instead of managing countless independently deployed agents, a unified platform can simplify governance and maintenance, allowing IT to focus on the software while business teams manage their workflows. 💬 How are you handling governance and maintenance for private AI agents in your current pipeline? Let's discuss in the comments. 👇 Want to explore how private AI agents can move from prototypes to production? Join us at MLcon Berlin 2026! 📍 MLcon Berlin | Maritim ProArte Hotel Berlin | November 16–20, 2026 mlconference.ai/berlin/progr… 🎤 Explore sessions on AI Agents & Agentic Workflows, MLOps, Open Source LLMs, and AI Governance. 🎫 Last tickets available! MLcon Berlin is almost here — don't miss your chance to join us! 🚀 mlconference.ai/berlin/ticke…

🚀 How do you run AI agents in parallel without sacrificing reliability? Parallel tool execution can make multi-agent systems more efficient. But what happens when agents make decisions based on different assumptions? 🎙️ Francesco Esposito shares a practical approach: Keep shared assumptions in a central Markdown or JSON file, giving agents visibility into each other's decisions. 💡 The key takeaway? Reliable multi-agent systems need shared context, not just parallel execution. Want to learn more? Meet Francesco at MLcon Berlin 2026, where he'll share practical insights in two sessions: 🔹 Designing Production-Ready Agentic Systems: Lessons from the Field mlconference.ai/ai-agents-ag… 🔹 Embodied AI: Vision-Language Models Meet Physical Systems mlconference.ai/machine-lear… 📍 Maritim proArte Hotel Berlin 📅 November 16–20, 2026 | Berlin or Online 👉 Discover more about Francesco and his talks: mlconference.ai/speaker/fran… 🎟️ Secure your ticket for MLcon Berlin: mlconference.ai/berlin/ticke… #MLcon #MLconBerlin #AIAgents #AgenticAI #AIEngineering

📌 What happens when AI agents write more code and developers become the conductors? 🎬 Insight from Apurva Misra, recorded live at MLcon. THE QUICK BREAKDOWN: 💡 AI-augmented engineering shifts developers toward higher-level responsibilities. As AI agents handle more tasks, developers increasingly guide decisions, provide context, and coordinate workflows. 💡 Business knowledge and cross-functional expertise become essential. Developers need to understand different domains, collaborate across teams, and answer the business questions AI agents cannot resolve independently. 💬 How is AI changing your role as a software developer? Are you becoming more of a coder or a conductor? Let's discuss in the comments. 👇 Ready to explore how AI is transforming software engineering? Join us at MLcon Berlin 2026! 📍 MLcon Berlin | Maritim ProArte Hotel Berlin | November 16–20, 2026 | mlconference.ai/berlin/progr… 🎫 Last tickets available! MLcon Berlin is almost here — don't miss your chance to join us! 🚀 mlconference.ai/berlin/ticke… #MachineLearning #AI #Tech #MLOps #MLcon #AIAugmentedEngineering #SoftwareDevelopment

What are an AI agent’s fingertips? Internal developer portals were built for humans, menus, click-through workflows, guided paths. But an AI agent doesn’t click through anything. So what happens when agents become regular platform users? In this short video, Natalia breaks it down: how to give agents the right capabilities, context, and controls, without rebuilding your whole platform around them. Her latest article goes deeper on exactly that. devopscon.io/blog/internal-d… #MLcon #AIAgents #PlatformEngineering #MLOps #AIEngineering #InternalDeveloperPlatform

📌 Your AI agent gets positive feedback. But what if every query costs 100x more than it should? 🎬 Insight from @Saumya Goyal, recorded live at MLcon. THE QUICK BREAKDOWN: 💡 Traditional ML models rely on metrics like F1 scores and accuracy. But evaluating RAG systems and AI agents requires more than a single performance score to determine production readiness. 💡 Observability reveals what's happening under the hood. An agent might deliver correct answers while consuming 20,000 tokens instead of 200 per query, creating hidden costs and operational risks. 💬 How are you evaluating AI agents and RAG systems before deploying them to production? Let's discuss in the comments. 👇 Want to learn how to evaluate AI agents and RAG systems before they reach production? 💬 How are you evaluating AI agents and RAG systems before deploying them to production? Let's discuss in the comments. 👇 Want to learn how to evaluate AI agents and RAG systems before they reach production? Join us at MLcon Berlin 2026! 📍 MLcon Berlin | Maritim ProArte Hotel Berlin | November 16–20, 2026 | mlconference.ai/berlin/progr… 🎫 Last tickets available! MLcon Berlin is almost here — don't miss your chance to join us! 🚀 mlconference.ai/berlin/ticke… #MachineLearning #AI #Tech #MLOps #MLcon #AIObservability #RAG

🚨 Your AI model works in testing. But can you prove it still works in production? A model can deliver excellent results today and quietly become unreliable tomorrow. Data shifts, performance degrades, and without proper monitoring, failures can go unnoticed. That's why MLOps is no longer just about deployment. It's becoming the trust layer of modern AI. In our latest article, @nahlawrites explores how MLOps helps teams build AI systems that are reliable, traceable, and accountable. 🔍 Here's what you'll discover: 🔹 Model Drift: Why AI performance can deteriorate silently and how continuous monitoring helps detect it. 🔹 Data Provenance: How to trace datasets, transformations, and predictions back to their origins. 🔹 Versioning Beyond Git: Why models, datasets, and experiments all need proper version control. 🔹 Production Reliability: How monitoring, model registries, and automated pipelines help teams maintain trustworthy AI systems. 💡 The real challenge isn't getting AI into production. It's keeping it reliable once it's there. 📖 Read the full article: mlconference.ai/blog/beyond-… #MLcon #MLconBerlin #MLOps #MachineLearning #AIEngineering

When people first made text readable for the blind, they embossed printed letters. Obvious. It barely worked, fingertips couldn’t tell the shapes apart. Then Louis Braille rethought it from scratch: six dots, built for touch, not sight. Same text, a representation designed for how it would actually be read. Giving AI agents access to your internal developer portal is the embossed-letters version. The portal was built for human fingers: menus, templates, golden paths. An agent doesn’t click through any of it. It needs the platform underneath, in a form software can consume. Based on what we learned at DevOpsCon New York, our latest article breaks down the three foundations of an agent-ready platform: Capabilities the agent can invoke, like provisioning infrastructure or deploying apps. Context about the systems those actions touch. Controls that define what the agent is and isn’t allowed to do. Read the full piece on how to implement this in your IDP. devopscon.io/blog/internal-d… This is exactly the conversation at DevOpsCon Munich, where Russell Miles leads the AI Platform Engineering Day roundtable, Hossein Salahi asks us to forget the portal entirely, and Matteo Emili, Max Körbächer, and Johannes Rudolph dig into what platform engineering becomes in the agent era. Explore AI Platform Engineering Day : devopscon.io/ai-platform-eng… #MLcon #AIAgents #PlatformEngineering #MLOps #AIEngineering #InternalDeveloperPlatform

🔍 From Traceable Answers to Trustworthy AI: What does it really take to build reliable AI systems? Building an AI demo is one thing. Making it reliable, measurable, and ready for production is another. At MLcon Berlin, industry experts share practical approaches to AI evaluation, agent reliability, and production-ready architectures. 🎤 Explore these four sessions: 🔹 The Failure-to-Metric Method: Building Evals Your Platform Can't Maya Malamud | Malamud AI Advisory 🔹 Building agent interfaces designing for a user that is a machine Michael Hablich | Google 🔹 From Demos to Durable Agents: Engineering the Reliability Stack for Production Shahid Malika Kandy | Nansen 🔹 Designing Production-Ready Agentic Systems: Lessons from the Field Francesco Esposito | Youbiquitous 💡 Because trustworthy AI isn't just about generating answers. It's about evaluating performance, designing reliable systems, and making AI work beyond the demo. 🚀 Ready to move from promising prototypes to production-ready AI? 👉 Explore the full MLcon Berlin program: mlconference.ai/berlin/progr… #MLcon #MLconBerlin #AgenticAI #AIEngineering #MLOps

💀 A fake skull fooled science for 40 years. Could your AI fool you in seconds? In 1912, the “Piltdown Man” was presented as a missing link in human evolution. For decades, scientists accepted the claim, until it was exposed as a hoax. The lesson? A convincing answer isn't necessarily a correct one. The same challenge exists with AI today. When retrieval fails to find the right information, even confident-sounding answers can be incomplete or misleading. 📘 In our latest FREE whitepaper, Ryan Paul Lafler, CEO of Premier Analytics Consulting, explains how to build a practical Retrieval-Augmented Generation (RAG) workflow using Python and LangChain. Here's what you'll learn: 🔹 Prepare and chunk documents for effective retrieval. 🔹 Use embeddings and FAISS to find relevant information. 🔹 Ground AI responses in approved, traceable sources. 🔹 Evaluate retrieval quality and identify gaps before they affect your answers. RAG doesn't guarantee accuracy. But it gives AI something better to work with: verifiable evidence. 📥 Download the FREE whitepaper: mlconference.ai/building-int… 🚀 Want to take your RAG systems further? Explore Advanced RAG, retrieval quality, and production-ready AI at MLcon Berlin. 👉 Explore the program: mlconference.ai/berlin/progr… #MLcon #MLconBerlin #RAG #GenerativeAI #AIEngineering

🎙️ 700 milliseconds. That's all the time a voice AI agent has to respond. But what happens when a tool call takes longer than that? In our latest MLcon Q&A, Francesco Esposito shares how his team tackles one of the toughest engineering challenges in voice-based AI: latency. 💡 Why tool-call reliability remains a challenge ⚡ How a two-level filler system keeps conversations flowing 🤖 How streaming responses can reduce perceived waiting time Because building production-ready AI agents isn't just about intelligence. It's about making interactions feel natural, even when complex processes happen behind the scenes. 🎥 Watch the Q&A and discover the engineering challenges behind real-time AI agents! 👉 Learn more about Francesco Esposito: mlconference.ai/speaker/fran… #MLcon #AIAgents #VoiceAI #AgenticAI #AIEngineering

Huge thanks to NVIDIA and Jensen Huang for flying into Berlin to be our opening act.  GTC is fun if you want to look at giant silicon wafers, talk about humanoid robots, and pretend your company is an AI lab. But unless your everyday users are factory robot arms, your biggest AI problem isn't compute. It's software. It's knowing how to build reliable agents, manage context, design MCP architectures, and keep your production systems from imploding when an LLM decides to freestyle. That’s why AI Native Week exists (Berlin, Nov 16–20). 4 conferences. 100+ sessions. Zero hardware sales pitches. Just pure, unadulterated software engineering across MLCon, API Conference, Vibe Kode, and the AI Engineering Summit. Let NVIDIA sell you the engine. We’ll teach you how to drive it without crashing into a wall. Grab your ticket: devm.io/ai-native-week/ #GTC26 #BerlinTech #AIEngineering #DevOps #Developers

📌 Is your enterprise AI exposing sensitive business data to external LLM providers? 🎬 Insights from @schmichri and Tim Frey recorded live at MLcon. THE QUICK BREAKDOWN: 💡 Connecting enterprise databases to external LLMs can expose sensitive business information to third-party AI providers. 💡 Data privacy starts with architecture. Enterprises need to understand how their data flows, where it's processed, and who has access to it. 💬 How are you protecting sensitive business data when integrating LLMs into your enterprise pipelines? Let's discuss in the comments. 👇 Want to explore how to build secure, enterprise-ready AI systems? Join us at MLcon Berlin 2026! 📍 MLcon Berlin | Maritim ProArte Hotel Berlin | November 16–20, 2026 | mlconference.ai/berlin/progr… 🎫 Last tickets available! MLcon Berlin is almost here — don't miss your chance to join us! 🚀 #MachineLearning #AI #Tech #MLOps #MLcon #DataPrivacy #EnterpriseAI

Is your machine learning model underperforming? The problem might not be the model itself. 🤔 Sometimes, better results come from adjusting how you train your model, not from building a more complex one. In our latest MLcon article, Arno Haase explores the practical side of hyperparameter tuning and explains how small decisions can make a big difference in ML training. 🔍 Here's what you'll learn: ⚙️ Learning Rate: Why choosing the right step size matters for training speed and convergence. 📊 Batching: How to train efficiently on large datasets without processing everything at once. 🧠 Model Selection: Why more complex models aren't always the better choice. 🎯 Input Parameters: How domain knowledge and better feature selection can improve model performance. The key takeaway? Better machine learning isn't just about training harder. It's about making smarter choices. 👉 Read the full article and explore practical approaches to improving your ML models: mlconference.ai/blog/hyperpa… #MLcon #MachineLearning #HyperparameterTuning #MLTraining #DataScience

🔐 Scaling AI in your enterprise? But how do you keep it secure, governed, and connected to your systems? At MLcon Berlin 2026, join Stefan Asanin, Senior Solutions Engineer at Axway, for the hands-on TechLab “Design, Secure, and Govern MCP Servers.” This isn't just theory. You'll get hands-on experience with: 🔹 Building a functional MCP server and connecting it to a dynamic MCP registry. 🔹 Routing AI interactions through an MCP gateway. 🔹 Managing authentication, security, and quotas. 🔹 Orchestrating data flows and maintaining observability across AI interactions. 💡 One endpoint. Full discovery. Governed AI connectivity at scale. 📅 Wednesday, November 18, 2026 🕒 15:00–18:00 📍 MLcon Berlin | TechLab ✅ Already have a Wednesday pass? This workshop is included — no separate booking required! 🔗 Explore the workshop: mlconference.ai/ai-agents-ag… 🎟️ Explore MLcon Berlin & Tickets: mlconference.ai/berlin/ticke… #MLcon #MLconBerlin #MCP #AIAgents #AIGovernance

📌 What if your AI coding agent never had to guess what you meant? 🎬 Insight from @Daniel Sogl recorded live at MLcon. THE QUICK BREAKDOWN: 💡 Eliminate ambiguity with detailed specifications. Instead of letting an AI agent guess whether a debounce search should use 300ms or 500ms, define the exact behavior upfront. 💡 Turn specifications into strict implementation guidelines. When feature requirements are fully documented, AI agents can focus on execution rather than making assumptions. 💬 How are you handling specification clarity in your AI-assisted development pipeline? Let’s discuss in the comments. 👇 Get more hands-on experience at MLcon Berlin 2026! 📍 MLcon Berlin | Maritim ProArte Hotel Berlin | November 16 – 20, 2026 | mlconference.ai/berlin/?utm_… 🎫 Join us in Berlin or online and discover practical insights into AI, Machine Learning, and MLOps! #MachineLearning #AI #Tech #MLOps #MLcon #AICoding #SpecDrivenDevelopment

⏰ LAST CALL: MLcon Berlin Early Bird ends TONIGHT! Your last chance to save up to €310 on your MLcon Berlin ticket is here! Five days of practical insights into Generative AI, AI Agents, RAG, MLOps, and AI Governance — focused on turning AI ideas into production-ready solutions. 🎟️ Here's what you can still secure today: 💰 Save up to €310 with Early Bird pricing. 👥 Get 10% off when booking 3 or more tickets. 🎥 Enjoy 1 year of devmio Fullstack Access + 6 months of conference recordings with qualifying passes. 🚀 Upgrade to AI Native Week for €299 and explore even more expert sessions across four conferences. 📍 Berlin | November 16–20, 2026 ⚠️ Early Bird offers end TODAY, October 15! Don't wait until prices go up. Secure your ticket before the clock runs out! ⏳ 👉 mlconference.ai/berlin/ticke… #MLcon #MLconBerlin #EarlyBird #AIEngineering #MLOps

⏰ ONLY 48 HOURS LEFT! Save up to €310 on MLcon Berlin tickets! AI demos are easy. Production isn't. 🚀 From AI agents and Advanced RAG to MLOps, multimodal AI, and governance, MLcon Berlin 2026 brings together the expertise you need to build AI systems that work beyond the prototype. Here's what you can look forward to: 🚀 Practical AI insights: Explore production-ready AI agents, LLMOps, open source LLMs, and more. 🛠️ Hands-on learning: Join Bootcamps, Power Workshops, and the AI Product Validation Day. 🎥 Extended learning: Get 1 year of devmio Fullstack access + 6 months of video recordings with qualifying passes. 👥 Team savings: Get an additional 10% discount when booking 3 or more tickets. 💥 Early Bird ends October 15! Save up to €310 before ticket prices increase. 📍 Berlin | November 16–20, 2026 ⏳ The clock is ticking! Don't miss your chance to learn, connect, and level up your AI engineering skills. 🎟️ Secure your ticket now: mlconference.ai/berlin/ticke… #MLcon #MLconBerlin #EarlyBird #AIEngineering #MLOps