Internal developer platform
A curated self-service layer that lets product teams build and operate software without mastering every infrastructure subsystem.
ConceptAdopt
Platform engineering knowledge graph
Platform engineering, mapped.
Explore the systems, practices, and tools that turn infrastructure into a dependable product.
GRAPH SCOPE
95 nodes
7 operating clusters, from developer experience to AI governance. Curated for production teams, not alphabetical completeness.
How to use this graph
Find a signal
Search aliases, sources, tools, or practices. Filters and production journeys narrow the graph and index together.
Inspect the node
Select any node for a practical definition, operating context, adoption signal, and deeper explanation.
Trace relationships
Arrowed, typed lines explain what enables, governs, secures, observes, or measures the selected node.
Trace a production journey
All platform capabilities and controls.Relationship map
Follow the operating system
Tab into the map, then use the arrow keys, Home, or End to move between nodes; each one is inspected as it receives focus. Drag to pan and use the zoom controls, or Control or Command with the scroll wheel, to change scale. Edges point from the selected node toward related capabilities and use line style to identify the relation type.
How to read the relation lines
EnablesImplementsDepends onGovernsObservesSecuresMeasuresAlternative toPart ofOften confused with
Searchable index
Every node, readable without the map
Platform as a product
Operating an internal platform with defined users, outcomes, discovery, roadmap, and adoption metrics.
PracticeAdopt
Golden paths
Supported, opinionated routes for common tasks such as creating, deploying, and observing a service.
PatternAdopt
Paved roads
Well-supported technology choices that teams are encouraged, but not always forced, to use.
ConceptAdopt
Developer portal
A searchable interface for ownership, documentation, templates, dependencies, and operational actions.
PatternTrial
Backstage
An open-source framework for developer portals, software catalogs, documentation, and templates.
ToolTrial
Self-service
Letting engineers complete approved platform actions on demand through automated interfaces.
PracticeAdopt
Cognitive load
The information and decision-making capacity engineers must hold to deliver and operate a system.
ConceptAdopt
Team Topologies
A model for organizing stream-aligned, platform, enabling, and complicated-subsystem teams around fast flow.
ConceptAdopt
Service catalog
A structured inventory of software components, owners, dependencies, lifecycle state, and operational metadata.
PatternAdopt
Terraform and OpenTofu ecosystem
Two distinct declarative infrastructure-as-code projects that share HCL workflows and much of the provider ecosystem.
ToolAdopt
Pulumi
Infrastructure as code using TypeScript, Python, Go, C#, and other general-purpose languages.
ToolAssess
Crossplane
A Kubernetes control-plane framework for composing and continuously reconciling cloud resources.
ToolAssess
Landing zones
Preconfigured cloud foundations for identity, networking, logging, policy, account structure, and cost controls.
PatternAdopt
Policy as code
Expressing and testing governance rules in versioned, executable policy.
PracticeAdopt
Immutable infrastructure
Replacing deployed infrastructure artifacts instead of modifying them in place.
PatternAdopt
Drift detection
Finding differences between declared infrastructure or configuration and actual running state.
PracticeAdopt
Secrets management
Controlling the creation, storage, distribution, rotation, and revocation of credentials.
PracticeAdopt
Infracost
A tool that estimates cloud cost changes from Terraform and OpenTofu plans.
ToolTrial
GitOps
Operating declarative systems from versioned, immutable desired state that software agents automatically pull and continuously reconcile.
PracticeAdopt
Argo CD
A Kubernetes GitOps controller with application health, synchronization, drift detection, and rollback visibility.
ToolTrial
Flux CD
Composable Kubernetes GitOps controllers for sources, manifests, Helm releases, and image automation.
ToolTrial
Progressive delivery
Releasing changes to increasing portions of traffic while evaluating health and business signals.
PracticeAdopt
Canary release
Sending a small share of traffic to a new version before wider promotion.
PatternAdopt
Rollback
Restoring a known-good application, configuration, model, or infrastructure state after a failed change.
PracticeAdopt
Software supply-chain security
Protecting the integrity and provenance of source, dependencies, builds, artifacts, and deployment inputs.
PracticeAdopt
Software bill of materials
A machine-readable inventory of software components and dependencies in an artifact.
PracticeAdopt
Kubernetes
A declarative orchestration system for scheduling and operating containerized workloads across clusters.
ToolContext dependent
Container Network Interface
A standard interface used by container runtimes and Kubernetes to configure workload networking.
ConceptContext dependent
Service mesh
A dedicated layer for service-to-service traffic policy, identity, telemetry, and resilience.
PatternAssess
Kubernetes operators
Controllers that encode domain-specific lifecycle logic behind Kubernetes custom resources.
PatternTrial
Autoscaling
Adjusting workload capacity in response to demand, schedules, or predictive signals.
PracticeAdopt
GPU scheduling
Allocating scarce accelerator capacity with placement, sharing, priority, and quota controls.
PracticeTrial
KServe
A Kubernetes-native model serving system with standardized inference resources, autoscaling, and rollout capabilities.
ToolAssess
Workload identity
Giving software workloads short-lived, verifiable identities instead of static credentials.
PatternAdopt
Observability
The ability to understand a system’s internal state from the telemetry and context it emits.
ConceptAdopt
OpenTelemetry
A vendor-neutral standard and toolkit for producing, collecting, and exporting traces, metrics, and logs.
ToolAdoptFast moving, reviewed 31 Jul 2026
SLI and SLO
An SLI measures user-relevant behavior; an SLO defines the acceptable target over a time window.
PracticeAdopt
Error budget
The allowed amount of unreliability implied by a service level objective.
ConceptAdopt
Distributed tracing
Following a request through multiple services and operations using shared trace context.
PracticeAdopt
Incident response
A coordinated process for detecting, containing, resolving, and learning from production disruption.
PracticeAdopt
Toil
Manual, repetitive, automatable operational work that scales with service growth and creates little enduring value.
ConceptAdopt
Model registry
A controlled inventory of model versions, metadata, lineage, evaluation results, and promotion state.
PatternAdopt
MLflow
An open-source platform for experiment tracking, model packaging, evaluation, and registry workflows.
ToolAdopt
Feature store
A system for defining, discovering, computing, and serving reusable machine-learning features.
PatternTrial
Evaluation gates
Automated criteria that block model, prompt, or retrieval changes unless quality and risk checks pass.
PracticeAdopt
LLM gateway
A controlled ingress for model calls that centralizes routing, authentication, budgets, policy, and telemetry.
PatternAdopt
Retrieval-augmented generation
Grounding model responses with information retrieved from an external knowledge source at request time.
PatternTrial
LLM guardrails
Versioned controls that constrain model inputs, outputs, tool use, and policy-sensitive behavior.
PracticeAdopt
Vector store
A database or index optimized for similarity search over embedding vectors and metadata.
PatternTrial
vLLM
An open-source inference engine optimized for high-throughput serving of large language models.
ToolTrial
FinOps
A cross-functional practice that makes cloud and AI value, cost, and ownership visible and actionable.
PracticeAdopt
Cost attribution
Assigning infrastructure and service spend to accountable teams, products, environments, or workloads.
PracticeAdopt
FOCUS
The FinOps Open Cost and Usage Specification version 1.2, a normalized schema for billing data.
ConceptTrial
Identity and access management
Policies and systems that authenticate identities and authorize actions on resources.
PracticeAdopt
EU AI Act evidence
An engineering implementation pattern for collecting traceable artifacts about how an AI system was designed, evaluated, changed, monitored, and overseen.
PracticeAdoptFast moving, reviewed 31 Jul 2026
PII redaction
Detecting and removing or transforming personal information before it reaches unauthorized systems.
PracticeAdopt
Data lineage
A trace of how data moves and changes from source to model, report, or application.
ConceptAdopt
Governance as code
Encoding controls, evidence requirements, and decision rules into versioned automation.
PracticeTrial
Platform capability
A reusable outcome that the platform exposes to internal customers, such as deploying a service or obtaining a cloud account.
ConceptAdopt
Platform metrics
Measures that show whether platform capabilities improve developer flow, reliability, adoption, and satisfaction.
PracticeAdopt
Team API
An explicit contract describing a team’s ownership, responsibilities, services, communication paths, and interaction expectations.
PatternTrial
Platform API
A stable programmatic contract through which teams request and manage platform capabilities.
PatternTrial
Production readiness scorecard
A transparent set of checks that evaluates whether a service meets agreed operational standards.
PracticeTrial
Environment as a service
On-demand creation of policy-compliant, observable, and disposable application environments.
PatternTrial
Cloud account vending
Automated creation of cloud accounts, subscriptions, or projects from an approved landing-zone baseline.
PatternAdopt
Continuous integration
Frequent integration of small changes into a shared branch with automated build and verification feedback.
PracticeAdopt
Artifact provenance
Verifiable information about where, when, and how a software or model artifact was produced.
ConceptAdopt
Artifact signing
Cryptographically binding an identity to an artifact digest so consumers can verify who approved or produced it.
PracticeAdopt
Artifact attestation
A signed statement that makes a claim about an artifact, such as its provenance, test result, or policy review.
ConceptTrial
Dependency health
Assessing whether third-party components are maintained, secure, governable, and replaceable enough for production use.
PracticeTrial
Feature flags
Runtime controls that separate code deployment from feature exposure.
PatternAdopt
Kubernetes Gateway API
A role-oriented Kubernetes API for expressing service networking and traffic routing.
ToolTrialFast moving, reviewed 31 Jul 2026
Dynamic Resource Allocation
A Kubernetes API framework for requesting, preparing, and sharing specialized hardware through resource claims.
ConceptAssessFast moving, reviewed 31 Jul 2026
Kueue
A Kubernetes-native job queueing controller for quota, admission, priority, and resource sharing.
ToolAssessFast moving, reviewed 31 Jul 2026
eBPF networking
Using Linux eBPF programs for programmable networking, security policy, and telemetry close to the kernel.
PatternTrial
Cluster multi-tenancy
Safely sharing Kubernetes cluster resources among teams or workloads with different trust and isolation needs.
PracticeAssess
Burn-rate alerting
Alerting on how quickly a service consumes its error budget across multiple time windows.
PracticeAdopt
Telemetry pipeline
The collection, processing, routing, redaction, sampling, and export path for traces, metrics, and logs.
PatternAdopt
Tail sampling
Choosing whether to retain a trace after enough of the complete trace is available for a policy decision.
PracticeTrial
Chaos engineering
Controlled experimentation that tests system behavior under realistic failure conditions.
PracticeTrial
Disaster recovery
Restoring critical services and data after a large-scale or long-duration failure.
PracticeAdopt
Prompt registry
A versioned system of record for prompts, templates, model settings, evaluation evidence, and promotion state.
PatternTrial
Inference gateway
A traffic-management layer optimized for routing inference requests across model-serving backends.
PatternAssessFast moving, reviewed 31 Jul 2026
Model monitoring
Continuous measurement of model quality, drift, operational health, and business impact after deployment.
PracticeAdopt
Data quality gates
Automated checks that block data or model pipeline progression when data contracts or quality thresholds fail.
PracticeAdopt
Agent harness
The runtime control layer around an AI agent’s model calls, tools, state, approvals, evaluation, and telemetry.
PatternTrial
Unit economics
Measuring technology cost against a meaningful product unit such as an active customer, successful task, or inference.
PracticeAdopt
AI system inventory
An owned register of AI systems, models, purpose, risk context, data, providers, dependencies, and lifecycle state.
PracticeAdopt
Human oversight
Designed points where qualified people can understand, intervene in, override, or stop an AI-supported process.
PracticeAdoptFast moving, reviewed 31 Jul 2026
Post-market monitoring
Ongoing collection and review of production performance, incidents, complaints, and emerging risks after an AI system is deployed.
PracticeAdoptFast moving, reviewed 31 Jul 2026
MCP governance
Controls for approving, authenticating, scoping, monitoring, and retiring Model Context Protocol servers and tools.
PracticeAssessFast moving, reviewed 31 Jul 2026
Cost anomaly management
Detecting, assigning, investigating, and correcting unexpected changes in cloud or AI spend.
PracticeAdopt
Software Carbon Intensity
A specification for expressing the carbon emissions rate of a software system per functional unit.
ConceptAssess
OpenLineage
An open standard for emitting lineage events about jobs, runs, datasets, and their relationships.
ToolTrial
The map is a system, not a stack ranking.
Adoption signals show a sensible default for mid-market production teams. Your constraints, workload, and operating capacity still determine the right route.
Platform foundations
16 nodes. Product, team, and experience principles that make a platform useful.
Infrastructure and IaC
10 nodes. Repeatable environments, policy boundaries, identity, and infrastructure lifecycle.
Delivery and GitOps
14 nodes. Safe, observable paths from source control to production.
Kubernetes and runtime
13 nodes. Workload scheduling, networking, scaling, and production execution.
Observability and SRE
12 nodes. Signals, objectives, and operating practices for dependable systems.
MLOps and LLMOps
14 nodes. Release, evaluate, serve, and govern models and AI applications.
FinOps, security, and governance
16 nodes. Cost, access, evidence, and controls that keep platforms accountable.