Browse Artificial Intelligence Community (217)

yairgil explains how the Azure Copilot Observability Agent in Azure Monitor helps teams investigate AKS incidents by correlating metrics, logs, traces, Kubernetes events, and change history into an evidence-backed root-cause narrative with recommended next steps.

azinh17 breaks down how Azure achieved a top MLPerf Training v6.0 result for Llama 3.1 405B, training at extreme scale across 8,192 GPUs. The post focuses on the cluster and network architecture choices—NVLink scale-up domains, Azure’s MRC fabric, and topology-aware parallelism mapping—that kept step time stable as the system scaled.

Anavi Nahar rounds up Azure Databricks announcements and sessions from Databricks Data + AI Summit 2026, focusing on tighter interoperability with Microsoft’s data stack (OneLake, ADLS) and governed access via Unity Catalog, plus new integrations like the Excel add-in, SharePoint ingestion, and OneLake catalog federation.

Alberto_Manuel explains why telecom operators need an ontology (semantic) layer to keep data meaning intact for GenAI and analytics, and outlines how Microsoft Fabric IQ (preview) uses ontology items, graph relationships, and data agents to enable cross-domain reasoning, governance, and scalable AI-driven data access.

yalavi explains how the Azure Copilot observability agent runs “deep investigations” to troubleshoot incidents by correlating telemetry across application, infrastructure, and platform layers, and by producing an evidence-backed narrative with clear mitigations rather than a single best-guess answer.

GeertVanTeylingen outlines a zero-copy pattern for making enterprise file data usable by modern AI and analytics platforms, using Azure NetApp Files as the system of record and Microsoft OneLake shortcuts to expose that data without migration or duplication.

GeertVanTeylingen explains how to build an enterprise RAG “knowledge pipeline” that can index and retrieve file-based content in place (no copy/migration) using Microsoft OneLake, Azure AI Search, and Azure OpenAI for embeddings and grounded answers with citations.

kinfey shows how to build a cloud-native evaluation harness for Azure AI Foundry skills using Foundry Hosted Agents, combining deterministic validators, an LLM judge that returns structured JSON, and a multi-turn adversarial attacker to catch regressions and compare models side by side.

RohitMadhavKrishnan introduces ArchAngel, an educational AI coding assistant designed to bring a team’s engineering standards directly into the IDE, so junior developers get constructive feedback while they write code. The post outlines the core idea, a reference architecture, and the Microsoft-centric stack used to ground guidance in “golden repos.”

BhaktiRath95 walks through common failure modes when running AI/ML inference workloads on Azure Container Apps, including slow model startup, probe timeouts, OOM kills, and GPU initialization problems. The post provides concrete probe settings, Python/FastAPI patterns, and Log Analytics queries to diagnose and fix issues methodically.

j_folberth explains how to deploy Azure AI Foundry Hosted Agents directly from a source-code ZIP instead of a container image, including the deployment lifecycle, an azd-based workflow, and a reusable GitHub Action that posts to the Foundry data plane and polls until the new agent version becomes active.

Heather Poulsen shares an optimization playbook for running agentic AI workloads in production on Azure, focusing on keeping multi-agent orchestration reliable while controlling token costs and latency. It highlights practical techniques like inference routing, prompt compression, RAG tuning, caching, and FinOps-style capacity planning.

Heather Poulsen outlines a governance-first blueprint for building scalable agentic AI systems, focusing on how to embed consistent controls and quality checks across user interactions, agent orchestration, integrations, data, and models so systems can scale without losing trust and oversight.

Heather Poulsen shares an event session overview on designing Azure AI Landing Zones as a production-ready foundation for deploying AI applications and AI agents at scale, with guardrails for networking, identity, security, governance, and cost control using Microsoft’s recommended architecture frameworks.

brauerblogs announces a two-day “Path to Production for Agents” webinar series (July 27–28) focused on moving agentic AI from prototypes to production, covering governance, landing-zone architecture, AgentOps practices, security risks like prompt injection, and cost/performance optimization with Azure Monitor and Microsoft Foundry.

Mayunk Jain summarizes the Azure App Service announcements from Microsoft Build 2026, including a new “Easy AI experience” with built-in MCP, GA of Isolated v4 for App Service Environments, and Managed Instance improvements for modernizing legacy apps (including IIS) with better diagnostics and deployment workflows.

jordanselig announces a public preview feature that lets Azure App Service expose an existing REST API as a Model Context Protocol (MCP) server using only an OpenAPI spec. The post covers how the platform generates MCP tools, how to configure it, and what to consider for authentication and safe exposure.

amolravande explains how to run agent-generated Python safely by combining Agent Governance Toolkit (AGT) policy enforcement with Azure Container Apps Sandboxes, using per-session microVM isolation plus a fail-closed egress proxy to reduce the blast radius of untrusted code.

kinfey explains how to run LLM agents that write and execute code without giving them a host-sized blast radius, using a MicroVM sandbox. The post walks through a real pipeline (a daily Mandarin World Cup podcast) built with Microsoft Agent Framework, Azure AI Foundry, and Hyperlight snapshot/restore isolation.

leoyao summarizes the //build 2026 updates to Foundry Toolkit for VS Code, focusing on an end-to-end Hosted Agent workflow (scaffold, run, deploy, observe), richer Toolbox integrations, and new LangGraph samples that cover MCP, human-in-the-loop flows, and production observability.

Ram Kakani explains how Oracle Managed Database MCP (Model Context Protocol) remote servers can be used from Microsoft Foundry to build enterprise AI agents that query Oracle AI Database@Azure, including local VS Code workflows, self-hosted Azure deployments, and a fully managed OCI option with identity, networking, and governance controls.

LZhang lays out a practical DevOps loop for Microsoft Foundry Hosted Agents, covering how to move from Terraform-provisioned infrastructure to production delivery with immutable agent versions, evaluation as a release gate, manifest-driven promotion, traffic-split canaries, and per-version observability.

madhurinrao introduces Azure Copilot Migration Agent, a guided workflow in Azure Migrate that connects discovery, assessment, planning, and execution for storage migrations—covering SMB/NFS file shares to Azure Files and Azure Blob container-to-container transfers.

j_folberth shows how to deploy a new version of an Azure AI Foundry Hosted Agent using a repeatable GitHub composite action, including required workflow prerequisites, action inputs/outputs, and a Bash + Azure CLI + jq approach for calling the Foundry data-plane REST endpoint safely.

kinfey breaks down a cost- and security-aware blueprint for running a multi-agent SDLC “tower” on AKS, using AI Runway for in-cluster model serving, Kata MicroVM isolation for each agent pod, and MCP so GitHub Copilot Chat can orchestrate tools while keeping token spend predictable.

MattMc announces new Azure Monitor capabilities for observing AI agents, including faster telemetry ingestion, larger event payloads for prompts/responses, an Agents fleet view, deeper end-to-end transaction debugging, and evaluation workflows (including human-in-the-loop annotations) across different hosting environments and frameworks.

j_folberth walks through what it takes to deploy an Azure AI Foundry Hosted Agent using the Foundry Service REST API, including the required Azure resources, container build/push flow, and the RBAC and managed identity setup needed for the Foundry project to pull images and create agent versions.

bobmital introduces Anyscale on Azure, an Azure Native way to run the Ray distributed runtime on AKS so teams can unify data prep, training, tuning, and serving in one system. The post focuses on architecture (split control/data plane), GPU utilization and scheduling features, and Azure-native identity, networking, and governance.

budzynski outlines new AI gateway features in Azure API Management, including a Unified Model API (preview) that standardizes clients on OpenAI Chat Completions while APIM translates to different model providers. The post also covers GA support for Anthropic/Vertex AI, richer token metrics in Application Insights, and expanded content safety for MCP and A2A traffic.

Priyanka Nanda summarizes the Build 2026 updates for Azure Monitor, including new agent observability features, the Azure Copilot Observability agent, expanded OpenTelemetry/OTLP ingestion, and improvements to alerts, metrics querying, and SLI/SLO tracking across services like AKS and Application Insights.

nzthiago summarizes the Build 2026 wave of Azure Functions updates, covering a new serverless agents runtime, first-class managed connectors, MCP improvements, refreshed local tooling (Functions CLI v5 and VS Code templates), Go support on Flex Consumption, Durable Task Scheduler enhancements, and new operational/security features like built-in Grafana dashboards and TLS certificates.

DivSwa introduces Azure Logic Apps Automation (public preview), a new SaaS-style SKU for building and running workflow automations on Azure with built-in governance and production controls. The post highlights AI-assisted authoring, agent integration options (including Foundry agents and GitHub Copilot harnesses), and enterprise features like VNet/private endpoints, RBAC, and audit logging.

DivSwa announces the public preview of Knowledge as a Service (KBaaS) in Azure Logic Apps, a managed knowledge layer that turns documents into a ready-to-use knowledge base for agentic workflows, removing the need to build and operate a custom RAG pipeline, vector store, and retrieval logic.

DivSwa announces an improved integration between Azure AI Foundry and Azure Logic Apps, aimed at running agents inside real workflows. It covers creating or invoking Foundry Agents directly from the Logic Apps designer, triggering agents from events or schedules, and exposing Logic Apps connectors and long-running workflows as agent tools.

lily-ma introduces Hosted MCP Servers in Azure Logic Apps Connector Namespace (public preview), a managed way to deploy remote MCP endpoints from a catalog so AI agents can discover and call tools without you owning the underlying infrastructure, scaling, authentication, or monitoring setup.

vyomnagrani announces the public preview of Azure Container Apps Sandboxes, a new Azure resource for fast, hardware-isolated, scale-to-zero compute that can suspend/resume via full-state snapshots. The post explains the resource model, lifecycle states, egress controls, managed volumes, identities, MCP connectors, and how to get started with the portal, aca CLI, and Python SDK.

Anthony Chu announces the Azure Functions serverless agents runtime (public preview), a markdown-first way to build and run AI agents as Azure Functions with triggers, scale-to-zero, and built-in operational integrations. The post outlines the .agent.md format, shared configuration files, MCP-based tools/connectors, and preview capabilities like sandboxed browser automation.

vyomnagrani summarizes the Build’26 updates for Azure Container Apps, focusing on new serverless primitives for agentic workloads: Sandboxes for fast, hardware-isolated ephemeral compute, Express for near-instant app provisioning, plus portal, security, and observability improvements for running production container apps.

WSilveira introduces Azure Connector Namespace (preview), a managed Azure integration layer that lets apps running on Functions, Container Apps, App Service, or self-hosted compute call connector actions and subscribe to triggers without owning auth, retries, polling, or webhook plumbing. The post also explains MCP servers for exposing connector operations as tools to Copilot and other agents.

beenamore summarizes the Azure Logic Apps announcements from Microsoft Build 2026, including a new Logic Apps Automation SKU, GA for the Logic Apps MCP Server, preview integrations with Microsoft Foundry agents, a “Knowledge as a Service” RAG capability, code-first workflows via the Logic Apps Standard SDK, and a migration agent for BizTalk Server.