Skip to main content

Attribute Service Description

Service overview​

Attribute™ delivers precise cost attribution for AI and cloud workloads. It uses eBPF to collect resource consumption data directly from the kernel. There are no code changes and no application instrumentation.

Shared infrastructure hides who consumes what. GPUs, clusters, and multi-tenant services blend many cost drivers into one bill. Attribute removes that blind spot. It maps spend to the team, workload, model, feature, environment, and customer that drove it.

A platform-led delivery model, supported by DoiT Forward Deployed Engineering (FDE), turns attribution data into decisions. You get granular cost signals and the engineering support to operationalize them.

Engagement model​

Platform-led attribution​

Attribute™ Platform – Granular AI and cloud cost attribution.

  • eBPF telemetry is collected at the kernel level. Collection is agentless to the application. No SDK and no code changes are required.

  • The attribution engine resolves consumption to dimensions you define: team, namespace, workload, model, feature, environment, and tenant or customer.

  • Real-time and historical views show how spend moves across those dimensions.

  • Coverage spans Kubernetes and multi-cloud compute, including shared GPU infrastructure.

  • Unattributed and untagged spend is surfaced so coverage gaps are explicit, not hidden.

Multi-cloud approach: Attribute provides cloud-agnostic attribution across major providers. This eliminates vendor lock-in and gives you one consistent cost model across your estate.

Forward Deployed Engineering (FDE) support​

DoiT Forward Deployed Engineers help you deploy the eBPF collector, validate attribution accuracy, define the attribution model, and integrate outputs into your FinOps workflows. FDE support is requested through the DoiT Cloud Intelligence platform via Expert Inquiry.

Account Manager​

An Account Manager acts as the main point of contact and owns the commercial and business needs.

Onboarding with DoiT​

DoiT’s onboarding approach is geared towards driving outcomes and value through the Attribute™ platform and the support of DoiT Forward Deployed Engineering.

Below is a representative timeline covering the first 90 days with DoiT.

During the first 60 days, you can expect to engage with the DoiT team on:

  • Onboarding Kickoff (Owner: DoiT Onboarding CSM): a live kickoff call, covering:

    • Team introductions
    • Your customized onboarding timeline
    • Key customer stakeholders involved (and recommended personas)
    • A review and validation of attribution goals discussed during the sales process
    • Accessing DoiT support and Forward Deployed Engineering
    • Initial platform setup: connecting cloud accounts and access roles, configuring single sign-on (SSO) with a supported identity platform, and planning eBPF collector deployment across your clusters
    • Meeting cadence, communication preferences, and the schedule for collector deployment and platform training

  • Environment Review and Collector Deployment (Owner: DoiT FDE): we review your cluster topology, kernel and OS compatibility, and security requirements, then deploy the eBPF collector and confirm telemetry is flowing.

  • Attribution Model Design (Owner: DoiT FDE): we define the attribution strategy that matches your business, including teams, models, features, environments, and tenants.

  • Platform Overview and Deep Dive (Owner: DoiT Onboarding CSM): a platform overview highlighting all features, followed by a second session to deep dive into advanced capabilities.

  • Use Case Specific Implementation (Owner: DoiT FDE): based on your environment and attribution model, your FDE team helps you address specific use cases (see Use case implementation for more detail).

  • Validation and Accuracy Review (Owner: DoiT FDE): we reconcile attributed spend against your cloud bills, quantify unattributed spend, and tune the model for accuracy.

  • Roadmap Outcomes and Retrospective (Owner: DoiT FDE): we review the initial 90-day period, provide two-way feedback, share learnings, and discuss tactical and strategic next steps.

  • Ongoing Enablement (Owner: DoiT FDE): your engagement with DoiT is about continuous value. We establish a regular cadence for syncs and reviews as your environment and attribution needs evolve.

Use case implementation​

As part of your engagement, DoiT Forward Deployed Engineering leverages what we learn about your environment to configure and implement real solutions using the Attribute™ platform, including:

  • Deploying eBPF collectors across clusters and accounts
  • Defining attribution strategy
  • Allocating shared GPU and infrastructure cost to the workloads that consume it
  • Building AI unit economics, including cost per inference, per model, per feature, and per customer
  • Configuring chargeback and showback reporting
  • Surfacing and reducing unattributed spend
  • Integrating attribution data into DoiT Cloud Intelligence dashboards, Allocations, budgets, and anomaly detection where present

Service components​

AI cost attribution​

Attribute spend to the workloads that generate it. eBPF telemetry captures CPU, GPU, memory, network, and storage consumption at the process and container level. The attribution engine resolves that consumption to teams, models, features, environments, and tenants. No application instrumentation is required.

Visibility and unit economics​

Turn attribution into unit economics. Measure cost per inference, cost per model, cost per feature, and cost per customer. Identify shared-cost leakage and unattributed spend. Track trends over time so cost moves with the decisions that drive it.

Customer responsibilities​

Deploy and maintain telemetry​

  • Customer will deploy and maintain the Attribute™ eBPF collector in the target environments, with DoiT FDE guidance.

  • Customer will run a supported operating system and kernel version. eBPF requires kernel compatibility, which DoiT will confirm during onboarding.

  • Customer will provide DoiT with the appropriate level of access to environments and cloud billing data to receive this service.

Timely cooperation​

Customer will respond promptly to reasonable requests for documentation or access to enable swift deployment and issue resolution.

Proactive collaboration​

  • Customer will work with DoiT Forward Deployed Engineering to keep the attribution model aligned with business goals as teams, models, and tenants change.

  • Customer will ensure proper coordination between internal platform, DevOps, and security teams to deploy collectors and act on provided guidance.

  • Customer will assign an Authorized Point of Contact who can provide approvals for commercial and financial decisions with DoiT.

Service exclusions​

Scope​

Human-delivered services do not include the following:

  • Direct execution of cloud infrastructure changes
  • Remediation or rightsizing of workloads. Attribute reports cost; it does not change running infrastructure
  • Operational incident support for cloud services. Attribute is not a replacement for existing Support plans with cloud providers
  • Application development and debugging
  • Custom instrumentation of application code. Attribution is delivered through eBPF, not code changes
  • Support for third-party observability or APM tools
  • Vendor contract negotiations unless specified under a separate agreement

Service availability​

Available globally, subject to regulatory restrictions and to operating system and kernel compatibility for eBPF collection.

Communication channels​

Customers can access recurring meetings with the DoiT account team based on a mutually agreed cadence. You can use Expert Inquiry to connect with Forward Deployed Engineering for further guidance.

General​

DoiT may update this service description from time to time. Notice of material changes to the service description will be made available to you.

Last updated on