Browse All Machine Learning Content (244)

John Savill shares a short Azure update for 14th August 2026, covering a channel milestone plus a set of platform changes across App Service, AKS, networking, security, storage, Fabric, PostgreSQL, and AI—including a note about Gemini 3.7 Flash appearing in GitHub Copilot.

sukkaur announces a preview of SQL database skills for Skills for Fabric, aimed at giving AI coding agents workload-specific guidance to create, connect to, query, and manage SQL databases as part of end-to-end Microsoft Fabric solutions.

meghasony announces a preview feature that lets dbt jobs run directly against Microsoft Fabric Lakehouse, bringing transformation and testing workflows closer to OneLake-backed data. The post explains the Lakehouse-first use cases and the exact configuration steps to connect an existing or new Lakehouse target.

pankajarora announces general availability of Capacity Overview Events in Microsoft Fabric Real-Time Hub, focused on monitoring capacity conditions and reacting to them with near real-time alerts and automated actions.

sbaynes describes CARE-X, a Microsoft Research chest X-ray vision-language model that combines report generation with calibrated structured predictions and spatial grounding. The post explains the model architecture, a DAPO reinforcement-learning alignment stage, benchmark results, and a separate tool-augmented measurement pipeline for measurement-dependent findings like cardiomegaly.

ArshadAliTMMBA announces the general availability of Fabric Runtime 2.0, the Apache Spark-based execution foundation for Microsoft Fabric’s Data Engineering and Data Science experiences, positioned as a modern, high-performance runtime integrated across the Fabric ecosystem.

fredguis announces general availability of identity-based predicate filtering for SQL Audit Logs in Microsoft Fabric, aimed at reducing audit noise at enterprise scale by filtering audit events based on who is querying and what they’re allowed to see.

goupadhy announces the July 2026 on-premises data gateway release (version 3000.326), focused on keeping Power BI Desktop and the Power BI Service aligned for refresh/query execution, and on improving security, authentication, diagnostics, and operational visibility for enterprise deployments.

Vishnu Charan TJ explains how new integrations between Azure Blob Storage and the NVIDIA Dynamo stack can speed up LLM inference on AKS, focusing on faster model weight loading for cold starts and KV cache offloading to reduce time-to-first-token latency.

Bret Myers explains how to break down Microsoft Fabric Data Warehouse capacity consumption by combining the Capacity Metrics app with Query Insights, so you can attribute warehouse-level CU usage to specific queries, users, and workloads during a selected time window.

Paul_VicenteH explains why retrieval and reranking are separate stages in a RAG pipeline, then demonstrates hybrid retrieval (including RRF) and learned reranking across Azure AI Search, Azure SQL, PostgreSQL Flexible Server, and Azure Cosmos DB using the SQuAD dataset.

msft-shreyas introduces the preview of “Build agent with AI” for Microsoft Fabric Data Agent, an assistant that helps creators explore schemas, learn from query history, generate agent and data source instructions, and validate example queries against real data before applying configuration changes.

John Savill introduces the core differences between AI, machine learning, deep learning, and generative AI in a short “Tech in Five” explainer, giving a practical mental model for how these terms relate and where LLMs fit in.

This week in machine learning, Microsoft Fabric focused on making data and events easier to reuse across analytics and ML pipelines, from OneLake mirroring previews (Azure Monitor Log Analytics and AWS Glue-cataloged Iceberg) to more end-to-end streaming options with Change Event Streaming, Eventhouse, and clearer event architecture guidance. On the operations side, Fabric added practical safeguards and runbook improvements with Item Recovery becoming the default, a new VNet gateway evaluation engine preview, and scheduled User Data Functions for managed recurring logic. We also got concrete agent-building guidance with Microsoft IQ MCP endpoints across web, work, Fabric, and Foundry, and Microsoft Research released PRISM2 pathology foundation model weights on Hugging Face for research and benchmarking.

John Savill runs through the 7th August 2026 Azure update, covering changes across compute, networking, security, storage, SQL, and Databricks—highlighting VM defaults and retirements, new routing and proxy capabilities, Private Link over IPv6, and several Databricks and Azure SQL enhancements.

Yulia Turchin announces a Microsoft Fabric admin change that enables Item Recovery by default for tenants without an explicit setting, providing a built-in recovery window for supported item types and guidance on retention, restore workflows, and OneLake storage billing impact.

goupadhy announces a preview opt-in setting for VNet data gateways in Microsoft Fabric: the modern evaluation engine. The post explains what workloads it applies to (Mashup evaluations), why it matters for refresh and DirectQuery performance, and what metrics and rollout practices to use when validating it in your environment.

mk_sunitha introduces scheduled User Data Functions in Microsoft Fabric (Preview), showing how teams can run custom, parameterized business logic on a recurring cadence using the Fabric job scheduler, and then monitor outcomes and failures from the Fabric Monitoring Hub.

Scirqueira explains how to choose a medallion architecture pattern in Microsoft Fabric Data Warehouse by focusing on how much Spark you actually need, then mapping Bronze/Silver/Gold layers to either an all-in-one warehouse approach or a Lakehouse + Warehouse hybrid.

robece-msft explains how to choose between Business Events, Fabric Events, and Azure Events in Microsoft Fabric, focusing on how each “event pillar” fits different event-driven architecture needs and enterprise implementation patterns across the Fabric platform.

Microsoft Developer shares Shireesh Thota’s perspective on turning “data everywhere” into usable intelligence by bringing data together, adding context, and applying AI on top—framed as an architecture discussion rather than a single feature walkthrough.

MattH_OneLake announces preview mirroring features in Microsoft Fabric that let teams surface Azure Monitor (Log Analytics) tables and AWS Glue-cataloged Apache Iceberg tables in OneLake using a zero-copy, metadata-based approach for cross-domain and cross-cloud analytics.

robece-msft continues the Microsoft Fabric Business Events series with guidance on designing business event schemas that downstream consumers can reliably interpret and act on in event-driven architectures.

sbaynes reports on PRISM2, a pathology foundation model from Microsoft Research and Paige (Tempus) that links tissue images with diagnostic language from real pathology reports, aiming to reduce the need to rebuild separate models for each cancer-detection task.

nzagorac-ms announces a preview feature for SQL database in Microsoft Fabric that streams row-level changes (inserts, updates, deletes) into Fabric Eventstream or Azure Event Hubs, enabling near-real-time pipelines that connect operational data to analytics workflows inside Fabric.

Tzvia introduces Eventhouse in Microsoft Fabric and explains how it helps data engineers ingest, store, and analyze high-volume streaming event data in near real time, reducing the usual tuning and plumbing required for streaming workloads.

Pamela Fox shares the materials from the “Microsoft IQ Deep Dive with Python” livestream series, showing how to ground AI apps and agents using Web IQ, Work IQ, Fabric IQ, and Foundry IQ via MCP endpoints, plus code samples, slides, and write-ups for building and deploying Python agents.

This week's ML roundup focuses on turning ML work into repeatable, production-friendly systems. Microsoft outlined an API-driven GeoAI pipeline for generating GIS-ready vector layers from Earth observation imagery using the MARS model, while Microsoft Research shared Echoverse, a set of stateful synthetic environments with database-grounded verifiers to make computer-use agent training and evaluation more reliable. On the data platform side, Fabric and OneLake updates emphasized zero-copy access to telemetry and open table formats, stronger governance and outbound controls, and more operable real-time and Spark foundations that feed analytics and ML workflows.

nelsontam introduces Microsoft’s MARS (Map Autoregressive) model on Microsoft Foundry and explains how it converts satellite imagery into GIS-ready vector map data. The post outlines an end-to-end workflow with Microsoft Planetary Computer Pro, including STAC/COG ingestion requirements and practical deployment settings for running geospatial feature extraction at scale.

Samantha Kubota introduces Echoverse, a set of high-fidelity synthetic “worlds” for training computer-use agents, and explains why depth (stateful workflows and database-grounded grading) matters more than sheer environment count. The post shares results from supervised fine-tuning and reinforcement learning, plus links to released code and datasets.

John Savill rounds up a week of Azure platform updates, spanning AKS, networking, API Management, monitoring, data services, and regional expansion, plus a few notable AI model availability items and security-related changes like Azure Enclave and Key Vault symmetric keys.

diptiborkar summarizes why Microsoft Fabric and OneLake were recognized as a Leader in Forrester’s 2026 Data Lakehouses report, focusing on a unified lakehouse foundation for analytics and AI, open table formats, cross-cloud interoperability, and built-in governance and security across engines like Spark, SQL, KQL, and Power BI.

arindamc explains how Workspace Outbound Access Protection (OAP) in Microsoft Fabric controls outbound connectivity for Eventstream, so streaming pipelines can only read from approved sources and write to approved destinations (including cross-workspace routes) using data connection rules.

MattH_OneLake announces two Microsoft Fabric preview features that mirror Azure Monitor (Log Analytics) tables and AWS Glue Data Catalog-managed Apache Iceberg tables into OneLake. The post explains the zero-copy, metadata-based approach and outlines how to set up each mirroring source for analysis across Fabric workloads.

robece-msft outlines how to build event-driven applications in Microsoft Fabric using Business Events and User Data Functions, with code samples, architecture patterns, and practical scenarios aimed at teams building real-time, data-centric solutions.

Tzvia introduces Real-Time Dashboard in Microsoft Fabric, focusing on using streaming data to build live operational views that update in seconds so teams can spot spikes, slowdowns, and stalled processes quickly and respond while events are still unfolding.

ealroybrin introduces Microsoft Fabric Embed (Preview), which lets developers embed Fabric Real-Time Dashboards into JavaScript/TypeScript web apps using delegated Microsoft Entra ID authentication. The post outlines what the preview supports, key limitations (like no service principal auth), and how Fabric governance and permissions carry over to embedded experiences.

Santhosh_Ravin1 announces general availability of job concurrency and queue monitoring for Apache Spark in Microsoft Fabric, adding a capacity-level view to help diagnose why Spark jobs are queued or throttled and to understand how individual workspaces consume shared capacity over time.

Ilana Waitser introduces a preview capability that brings Azure Monitor telemetry into Microsoft Fabric so teams can connect operational signals (logs and metrics) with analytics workflows and understand business impact faster.

Ilana Waitser announces a public preview that lets you share Azure Monitor Logs from a Log Analytics workspace into Microsoft Fabric’s OneLake as Delta Parquet, without duplicating data. The post explains what this enables in Fabric—Power BI reporting, Spark-based analytics and ML, and correlating telemetry with business data for near real-time decisions.