Red Hat AI 3: A Unified Platform for Distributed LLM Inference

This week's AWS AI in Practice meetup highlighted the impactful role of AI-assisted software development and agents. I found it interesting that the focus is not just on the technology, but on practical applications that can reshape our workflows. What are your thoughts on how AI agents can transform our industry?

How do we build agents that can discover each other and exchange information remotely without REST APIs? because the more we wire these agents together with REST apis the more tangled they become. How do we build AI agents that can sit across the network and regions, ready to be scaled to any limit? What comes to mind is good old CORBA and RMI. These used to be popular in the old days for creating agents. Yes, ‘agents’ isn’t a new idea. Ive tried a flavour of CORBA with vert.X and tested with 3 virtual machines deployed in Azure and intentionally placed in different regions What you’d see is agents that are of aware of each other’s failure and agents that can be scaled horizontally. https://lnkd.in/dfSTS8bk

🚨🚨 𝗛𝗨𝗚𝗘 𝗡𝗘𝗪𝗦 for AI builders! Bedrock AgentCore is now GENERALLY AVAILABLE! 🚀 This is a game-changer for building production-ready AI agents. Here's what just dropped: ✨ 𝟴-𝗵𝗼𝘂𝗿 𝗲𝘅𝗲𝗰𝘂𝘁𝗶𝗼𝗻 𝘄𝗶𝗻𝗱𝗼𝘄𝘀 - Let your agents handle complex, long-running tasks 🔒 𝗩𝗣𝗖 𝘀𝘂𝗽𝗽𝗼𝗿𝘁 - Deploy agents securely in your private network 🛠️ 𝗦𝘁𝗿𝗲𝗮𝗺𝗹𝗶𝗻𝗲𝗱 𝗮𝗰𝗰𝗲𝘀𝘀 𝘁𝗼 𝘁𝗼𝗼𝗹𝘀 - Connect to MCP servers & transform ANY API into agent-compatible tools 🔐 𝗘𝗻𝗵𝗮𝗻𝗰𝗲𝗱 𝗜𝗱𝗲𝗻𝘁𝗶𝘁𝘆 - OAuth integration with secure vault storage & identity-aware authorization 📊 𝗙𝘂𝗹𝗹 𝗢𝗯𝘀𝗲𝗿𝘃𝗮𝗯𝗶𝗹𝗶𝘁𝘆 - End-to-end visibility with CloudWatch, plus integrations with Datadog, Dynatrace, LangSmith & more! The best part? Framework freedom! Use CrewAI, LangGraph, LlamaIndex, OpenAI SDK, or 𝗔𝗡𝗬 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 you love. Work with 𝗔𝗡𝗬 𝗺𝗼𝗱𝗲𝗹 inside or outside Bedrock. Zero infrastructure management. 💪 Available NOW in 9 AWS regions with consumption-based pricing (no upfront costs!) 🎯 Ready to build enterprise-grade agents? Check out the AgentCore Starter Toolkit on GitHub and dive into the blog! See the links to both in the comments section. ♻️ Repost this to share learning opportunities with your network and help others grow. 👋 Follow Ivan Kopas and AWS Training & Certification for more news, tips, and tricks to help you advance your skills and career. #AWS #GenAI #AIAgents #AmazonBedrock

Microsoft Releases '#Microsoft Agent Framework': An #OpenSource #SDK and Runtime that Simplifies the Orchestration of Multi-Agent Systems -“unifies core ideas from AutoGen (agent runtime and multi-agent patterns) with #Semantic Kernel (enterprise controls, state, plugins) to help teams build, deploy, and observe production-grade #AIagents and multi-agent workflows” Marktechpost AI Media Inc

🚀 Amazon Bedrock AgentCore is now generally available! Building production-ready AI agents just got significantly easier. AgentCore is a comprehensive platform that lets you deploy and operate highly capable agents at scale—without managing any infrastructure. What makes AgentCore powerful: ✅ Framework Freedom – Use CrewAI, LangGraph, LlamaIndex, OpenAI Agents SDK, or any custom framework. Your existing agent logic works as-is. ✅ Enterprise-Grade Platform – Built-in authentication, memory, observability, and security. Get VPC support, 8-hour execution windows, and complete session isolation. ✅ Zero Infrastructure Management – No servers or containers to manage. Just deploy your agents and scale automatically. ✅ Rich Ecosystem – Access to Code Interpreter, Browser automation, Gateway (with MCP server support), and comprehensive observability through CloudWatch, Datadog, Dynatrace, LangSmith, and more. What's new in GA: 1️⃣ VPC support for secure, private deployments 2️⃣ Agent-to-Agent (A2A) protocol support in Runtime 3️⃣ Model Context Protocol (MCP) server connectivity in Gateway 4️⃣ Identity-aware authorization with secure vault storage 5️⃣ OTEL-compatible observability with native CloudWatch integration Read the launch announcement here: https://lnkd.in/gn4n7BKQ The AgentCore Starter Toolkit makes getting started incredibly easy with quick starts for SDK development, deployment workflows, migrating existing Bedrock Agents, and Gateway integration. Ready to build your next AI agent? Explore the toolkit and dive deeper into AgentCore Runtime, AgentCore Memory, AgentCore Identity, and AgentCore Observability here: https://lnkd.in/giCkRcHW #AWS #AmazonBedrock #AIAgents #GenerativeAI #MachineLearning #AgenticAI

One region of AWS went down. How many CXOs realised that : 1. AI can’t replace humans 2. Do not fire your best employees because you hear AI buzzword 3. You must have a DR solution in place and if you already had it, you must do DR drills. 4. DevOps is irreplaceable 5. Chaos Engineering is a must 6. AI did not resolve the issue AWS had, humans did. 7. Network Tolerance is not an option. Recall CAP theorem 8. Automation of Infra is not an option, it’s must 9. 100% scalable and zero downtime systems are myths. 10. No system is perfect. I think None.

It’s been a big week for open source and developer tooling innovation: 💡 Microsoft just launched the Azure DevOps Local MCP Server, an on-prem bridge that lets AI assistants securely access DevOps context — think pull requests, work items, and test plans — without leaking data. 💡 Google unveiled the Coral NPU, an open-source RISC-V-based architecture designed for edge AI performance and privacy. It’s a step toward democratizing high-performance AI computing — one chip at a time. 💡 And BrowserStack dropped its Visual Review Agent, an AI-powered assistant that flags only meaningful visual changes in web tests, cutting through pixel-level noise. Together, these innovations show how open source and AI are reshaping how developers code, test, and deploy — faster, safer, and smarter. #OpenSource #GitHubInsights #DevOpsAI #CoralNPU #BrowserStack #RISCv

While working at Amazon Web Services (AWS), I came across Roo Code (which uses MCP - Model Context Protocol to connect multiple contexts for LLMs), and some cool automation work by John Capobianco. This got me thinking... why not integrate MCP into my network Lab? 🤔 So I built something fun using #containerlab to see how AI could actually help automate network operations. What I built: User → Gemini Agent → MCP Tools → pyATS → Network Devices (SSH) Basically, the agent uses MCP to control network automation tools on a 3-node Arista cEOS lab running in Containerlab. I made sure dangerous commands are blocked by design (safety first! ). You can list devices, run show commands, and push configs through natural conversation - pretty neat, right? Want to check it out? GitHub: https://lnkd.in/gW8hbths Quick YouTube Demo: https://lnkd.in/gEq6wV2p I'm also exploring ways to tackle LLM hallucination - experimenting with RAG, agentic workflows, config versioning, and J2lint. What's your take on AI in network automation? Would love to hear your thoughts! 💭 #NetworkAutomation #MCP #LLM #containerlab #AI #AWS #pyATS

🚀 Generative AI is everywhere… but how do we make it enterprise-ready? I just read SUSE’s article “The Role of Generative AI in Enterprise Innovation” and it nails the big picture: use cases, governance, security challenges, integration into existing stacks. Here’s my take: 👉 Innovation without trust is just a demo. Generative AI isn’t just about shiny models – it’s about securing the entire software supply chain. Traceability, provenance, reproducibility… these are the real foundations of enterprise adoption. StackState (acquired by SUSE) & Application Collection (apps.rancher.io) 👉 From Kubernetes to the Linux kernel: control matters. AI runs on clusters, runtimes, operating systems, and hybrid infrastructures. SUSE is one of the few vendors mastering the full stack – Kubernetes, Rancher PRIME, RKE2, Linux, and runtime protection with SUSE NeuVector aka SUSE Security. That’s how we build environments that are reliable and resilient. 👉 Automate security, not complexity. Security shouldn’t slow down delivery. By embedding controls, remediation, and policies as code into CI/CD, we ensure that innovation flows without compromise. 👉 Sustainable and sovereign innovation. At SUSE, we’ve been enabling enterprises with open source freedom and trust for 30+ years. Generative AI is just the next step: helping customers innovate while staying in control of their choices, their data, and their future. 🔗 Read the article here: The Role of Generative AI in Enterprise Innovation from Jen Canfor https://lnkd.in/dh4K7kCY #AI #GenerativeAI #OpenSource #Security #DevSecOps #SUSE Eric Lajoie Lionel Meoni Julien Blayes Thomas BRIOIS Pierre-Yves Albrieux Sébastien Carrias Anna Sapegina Martina Ilieva Thomas Di Giacomo Andreas Prins Jeff Price

Microsoft launched the Agent Framework, an open-source SDK that unifies Semantic Kernel and AutoGen into a single platform for building AI agents fast, with less code and enterprise reliability. It’s a strategic move to make Azure the default agent orchestration layer for business-ready, interoperable AI systems. Main Takes • Consolidates Microsoft’s agent stack into one open-source framework • Built on open standards (MCP, A2A, OpenAPI) for ecosystem interoperability • Natively integrates with Azure AI Foundry, Microsoft Graph, Redis, Elastic • Enables agent orchestration patterns (workflow, debate, reflection) with production controls • Supports both Python and .NET for enterprise deployment Microsoft is commercializing open-source orchestration, turning developer experiments into enterprise infrastructure. The Agent Framework positions Azure as the control plane for AI agents, giving businesses a ready-made foundation for building secure, observable, and extensible agentic applications across their data and workflow ecosystems.