Explore Ops
42 published engineering-operations disciplines, each with capabilities, practices, metrics, tools, sources, and a place to start.
Software Delivery & Engineering Ops
- DevOpsThe operating model where the teams who build software and the teams who run it share one lifecycle — measured by flow, feedback, and what customers actually experience.
- GitOpsGit as the single source of truth for declared state: agents pull, continuously reconcile reality against the declaration, and recovery is a revert.
- BuildOpsRunning the build system as a product: reproducible builds, correct caching, signed artifacts with known provenance, and a build time that does not tax every engineer on every change.
- ReleaseOpsGetting a built artifact to users deliberately: versioning, change approval, staged rollout, release communication, and a rollback that is exercised rather than assumed.
- TestOpsOperating the test infrastructure: environments and data that are available when needed, suites that finish quickly, and a flake rate low enough that a red build is believed.
- QAOpsThe quality function as an operating discipline: deciding what to test at which layer, running defect triage, analysing escaped defects, and setting the gates that a change must pass.
- DevExTreating the internal developer experience as a measured product: onboarding time, local setup, feedback-loop length, and the friction engineers actually report rather than the friction leadership assumes.
- ChatOpsConducting operations in the chat surface where the team already works: bots and commands that execute real actions, with authorization, audit and a shared transcript of who did what.
Reliability & Runtime Ops
- SREThe discipline that treats reliability as a software problem: SLOs define how unreliable you may be, error budgets turn that into policy, and engineering time replaces toil.
- Observability EngineeringInstrumenting systems so any novel question about production is answerable in minutes — traces, metrics, and logs unified by shared identifiers, priced sanely.
- IncidentOpsIncidents run as a practiced system: named roles, a severity taxonomy that activates automatically, honest customer comms, and a blameless loop that closes.
- Resilience EngineeringDesigning for the failure that will happen: no shared failure modes, graceful degradation, and rehearsals that prove recovery before production does.
- Chaos EngineeringDeliberately injecting failure as controlled experiments — a hypothesis, a bounded blast radius, a measurement and an abort condition — to learn how a system behaves before production teaches you for free.
- AIOpsApplying machine learning to IT operations signals — correlating events, suppressing noise and detecting anomalies — so that the humans on call spend attention on the few things that matter.
Cloud & Infrastructure Ops
- CloudOpsRunning a cloud estate as a governed, reproducible system: infrastructure as code, enforced tagging, guardrails, and deliberate Well-Architected trade-offs.
- Platform EngineeringThe internal platform run as a product: golden paths that make the right architecture the default, self-service for teams, and SLOs the platform answers for.
- ITOpsRunning IT as a service to the organisation: a service desk people can reach, request fulfilment with expectations, and change, incident and problem management applied to enterprise IT rather than to product engineering.
- SysOpsHands-on administration of running systems: patching on a cadence, configuration that is known rather than remembered, capacity headroom, host-level backups, and restores that have been tested.
- InfraOpsOwning the infrastructure estate as a defined, versioned artifact: provisioning through code, environment topology that is consistent, drift that is detected, and a clear picture of what exists and what it costs.
- NetOpsOperating the network as a dependable service: a current picture of topology and addressing, controlled change, monitoring that sees the path rather than the device, capacity planning, and response when the network is the incident.
- DevNetOpsApplying software delivery practice to networking: configuration in version control, automated validation before change, testable intent, and deployment pipelines for a domain where a bad change disconnects the estate.
- NoOpsAn honest account of the claim that operations disappears: what managed and serverless platforms genuinely remove, what they only move to a vendor, and what stays with you no matter what you buy.
Security, Risk & Governance Ops
- DevSecOpsSecurity and delivery as one pipeline: checks run at commit, build, and deploy time — with severity-based SLAs, not a review gate at the end.
- SecOpsSecurity operations: detecting, triaging and containing attacks as a run discipline — alert quality, on-call for security, and the drills that make response boring.
- CloudSecOpsCloud security operations: keeping posture, identity and workload configuration correct across accounts — guardrails that prevent drift instead of dashboards that report it.
- ComplianceOpsCompliance operations: turning control frameworks into continuously produced evidence, so an audit is an export rather than a three-week fire drill.
- GRCOpsGovernance, risk and compliance operations: running the risk register, policy lifecycle, exceptions and third-party review as living processes rather than annual documents.
- PrivacyOpsPrivacy operations: knowing what personal data you hold, deleting it when you should, and answering data-subject requests as a routine process rather than an emergency.
Data & Analytics Ops
- DataOpsThe operating model for data pipelines: getting data from where it is produced to where it is used, on time, unchanged in meaning, and with a known answer when it breaks.
- AnalyticsOpsThe operating discipline for the analytics consumption layer: metric definitions, the semantic layer, dashboard lifecycle, and making two reports agree about what revenue means.
AI & ML Ops
- MLOpsOperating the training-to-serving lifecycle of models you build: reproducible training, versioned data and features, tested deployment, and monitoring for the drift that degrades a model nobody touched.
- LLMOpsOperating applications built on large language models you mostly do not control: provider and version management, evaluation harnesses, cost and latency per request, guardrails, and a plan for the day the model changes underneath you.
- ModelOpsGoverning the whole portfolio of models an organisation is accountable for — including statistical and rules-based ones — through inventory, independent validation, approval, monitoring and retirement.
- PromptOpsTreating prompts as versioned production artifacts: a registry, change review, regression evaluation on a fixed test set, staged rollout and rollback — because a prompt edit is a behaviour change with no compiler to catch it.
- RAGOpsOperating the retrieval side of grounded AI systems: corpus ingestion and freshness, chunking and embedding versions, index rebuilds, retrieval quality, permission-aware search, and citations that actually support the answer.
Cost & Business Efficiency Ops
- FinOpsCloud spend as an engineering-owned business metric: allocate every dollar, know your unit cost, and run inform–optimize–operate as a continuous loop.
- BizOpsThe internal operating system of a company: the planning cadence, the business systems behind it, process design across functions, and the analysis that makes a decision defensible rather than loudest-voice.
Product & Go-to-Market Ops
- ProductOpsThe operational backbone of product management: customer feedback that reaches decisions, research operations, launch coordination, and product analytics instrumented well enough to answer the questions asked of it.
- RevOpsUnifying marketing, sales and customer success around one revenue process, one data model and one forecast — so the handoffs stop losing information and the number everyone reports is the same number.
- SalesOpsMaking the sales function run: a pipeline that reflects reality, CRM hygiene worth trusting, territory and quota design that holds, deal desk support, and forecasting that improves because it is scored.
- MarketingOpsThe execution layer of marketing: campaign operations, a martech stack that works together, lead lifecycle management, attribution that is honest about its limits, and consent handled as a real obligation.