Teaching a 9B model to investigate production alerts
Datadog has developed a "training-data flywheel" to make large-scale agentic incident investigations economically viable by fine-tuning a smaller, specialized model (Qwen3.5-9B) us…
Back to all articles
Feed
Datadog | The Monitor blog
1,919 articles · Page 3 of 160
When users don’t click thumbs up: Inferring agent feedback from Datadog telemetry
To address the lack of explicit user feedback for AI agents, Datadog researchers developed a "session classification skill" that uses weak labeling to approximate user satisfaction…
Cut AI agent cost and improve accuracy with Code Execution in the Datadog MCP Server
Datadog's new "Code Execution" feature enables AI agents to perform complex, multi-signal observability investigations using a sandboxed JavaScript environment. By allowing agents …
Understand the top paths users take to convert or drop off with Journey Paths
Datadog’s Journey Paths feature enhances traditional funnel analysis by visualizing the specific sequences of actions users take, revealing the actual behavior behind conversion an…
Enforce custom rules in Datadog IaC Security scanning
Datadog’s custom rules for IaC Security allow organizations to enforce specific internal standards, such as required tags or approved instance types, using Rego-based policies. The…
Transform and route security logs to Microsoft Sentinel tables using Observability Pipelines
Datadog Observability Pipelines’ Microsoft Sentinel Packs provide preconfigured mappings that convert various vendor-specific logs into standardized Sentinel schemas before they ar…
From alert to resolution: Manage incidents with Bits Chat in Slack
Datadog's Bits Chat integrates AI-driven incident response capabilities directly into Slack, allowing responders to investigate alerts, troubleshoot with telemetry data, and genera…
Manage Cursor costs with Datadog Cloud Cost Management
Datadog’s Cloud Cost Management (CCM) now integrates Cursor costs, providing engineering and FinOps teams with granular visibility into AI coding spend by user, model, and usage mo…
How to operate shared platforms safely at agent scale
To operate autonomous agents at scale, platform teams must move beyond basic design and prepare for new operational risks by modeling the entire "agent trajectory" to identify dive…
Monitor TAS and gang scheduling for AI training in Kubernetes
Standard Kubernetes scheduling fails to meet the complex requirements of distributed AI training, which demands both topology-aware placement for low-latency GPU communication and …
Datadog named the Company to Beat for observability platforms in 2026 Gartner® AI Vendor Race report
Datadog has been recognized by Gartner as a Leader in Observability Platforms and named the "Company to Beat" in the 2026 AI Vendor Race. To support the rise of AI-native applicati…
Understanding NetFlow duplication: Why it happens, and how to deduplicate
NetFlow duplication occurs when multiple network devices report the same traffic flow, resulting in inflated data that undermines accurate capacity planning and performance metrics…