---
title: "Engineering glossary: software, AI, cloud and hiring terms"
description: "Plain-language definitions of terms that come up when building apps, AI agents and cloud systems or hiring engineers, from caching to fractional CTO."
canonical: https://sdk.enterprises/en/glossary
language: en
---

# The terms behind the work, in plain language

Engineering conversations are full of jargon. These are short, plain definitions of the terms that come up most often when software, AI agents and cloud systems are built, run and staffed.

**API gateway**: A service that sits in front of one or more APIs and handles shared concerns such as routing, authentication, rate limiting and request logging. Clients call the gateway, which forwards each request to the right backend service.

**AWS**: Amazon Web Services, Amazon's cloud platform. It provides compute, storage, databases, networking and managed services on demand, replacing hardware a company would otherwise buy and run itself.

**Blue-green deployment**: A release technique that keeps two identical production environments. The new version is deployed to the idle one, checked, and then traffic is switched over from the live one. Switching traffic back is the rollback.

**Cache invalidation**: Removing or updating cached data when the underlying data changes, so users do not see stale results. Common strategies include expiry times, deleting entries on write and versioned keys.

**Caching**: Storing the result of an expensive operation, such as a database query or an API call, in fast storage so later requests can reuse it. Caching cuts response times and load on the underlying system, at the cost of keeping the cached data fresh.

**CI/CD**: Continuous integration and continuous delivery or deployment. Every change is built and tested automatically when it is merged, and a pipeline releases it to an environment with little or no manual work.

**Cloud migration**: Moving applications, data and infrastructure from on-premise servers or legacy virtual machines to a cloud provider such as AWS or Google Cloud. Approaches range from moving an application as it is to rebuilding it for managed cloud services.

**Code review automation**: Using static analysis, automated tests or AI agents to check code changes before or alongside a human reviewer. Automated checks catch style issues, common bugs and missing tests, so reviewers can focus on design and behavior.

**Cutover**: The moment a system switches from the old environment to the new one, for example when production traffic is redirected to a migrated application. A cutover plan lists the steps, owners, checks and rollback path.

**Data minimization**: The principle of collecting and keeping only the personal data needed for a stated purpose. It is one of the core principles of the GDPR, and it also limits the damage of any data breach.

**DevOps**: A way of working where the people who build software also automate how it is tested, deployed and run. In practice it means CI/CD pipelines, infrastructure described in code, and monitoring that tells the team when something breaks.

**Diff**: A line-by-line view of what changed between two versions of a file or codebase, showing added and removed lines. Diffs are the standard unit of code review.

**Elasticsearch**: A distributed search and analytics engine that indexes documents for fast full-text search and aggregations. It is often used for site search, log analysis and as a read-optimized layer next to a primary database.

**Engineering partner**: An external company that takes responsibility for an engineering outcome, not only for supplying hours. It plans the work, staffs it, answers for its quality and hands the result over to the client's team.

**Fractional CTO**: A senior technical leader who works with a company for part of the month instead of full time. They make technology decisions, help hire engineers and keep delivery on track, often for companies not yet ready for a full-time CTO.

**GDPR**: The General Data Protection Regulation, the European Union law that governs how personal data about people in the EU is collected, processed and stored. It applies to organizations that handle such data, wherever they are based.

**Google Cloud**: Google's cloud platform, also known as Google Cloud Platform or GCP. Like AWS, it provides compute, storage, databases, networking and managed services on demand.

**Idempotency**: The property of an operation that has the same effect whether it runs once or several times. Idempotent APIs let a client safely retry a request, such as a payment, after a timeout without creating duplicates.

**Kubernetes**: An open-source system for running containerized applications across a cluster of machines. It schedules containers, restarts failed ones, scales them up and down and manages rolling updates.

**Laravel**: A widely used open-source PHP web framework. It includes routing, an ORM, queues, caching, authentication and testing tools, and is a common choice for APIs and business applications.

**Legacy modernization**: Updating an older system so it is cheaper and safer to run and change. It can mean upgrading frameworks and language versions, restructuring code, moving to the cloud or rebuilding parts from scratch.

**Lift and shift**: A migration approach that moves an application to new infrastructure, usually the cloud, with little or no code change. It is fast, but the application gains little from managed cloud services until it is modernized.

**LLM agent**: Software that uses a large language model to plan and carry out multi-step tasks by calling tools, such as reading files, running tests or querying an API. Well-designed agents show their plan, log their actions and ask a human to approve changes.

**Observability**: The ability to understand what a system is doing from its outputs: logs, metrics and traces. Good observability lets a team find the cause of a production problem without first shipping new code to investigate it.

**OWASP**: The Open Worldwide Application Security Project, a nonprofit foundation that publishes free application security guidance. Its Top 10 list of critical web application security risks is a common baseline for secure development and code review.

**P95 latency**: The response time that 95 percent of requests are faster than. It shows what users experience on slower requests better than an average does, since an average can hide a long tail of slow responses.

**Rate limiting**: Capping how many requests a client can make in a given time window. It protects a service from overload and abuse, and requests over the limit are rejected or delayed.

**Real-time dashboard**: A dashboard that updates as new data arrives, usually over WebSockets or server-sent events, instead of on a schedule or a page refresh. Operations teams use them to watch systems and react to problems as they happen.

**Redis**: An in-memory data store used for caching, sessions, queues, rate limiting and real-time counters. Because the data lives in memory, reads and writes are very fast.

**Rollback**: Returning a system to its previous working version after a failed release or migration. A rollback path is planned and tested in advance so it can be carried out quickly and safely.

**Runbook**: A written procedure for operating a system: how to deploy it, check its health and respond to known alerts and failures. Runbooks let people other than the original authors run the system.

**Staff augmentation**: Adding external engineers to an in-house team to fill skill or capacity gaps, usually billed by time. The client directs the work day to day, and the provider supplies the people.

**Streaming ingestion**: Accepting and processing data continuously as it is produced, rather than in periodic batches. It suits network, sensor, event and log data where decisions depend on what is happening now.

**Team extension**: A way of adding specialists to an existing team for a defined period, working in that team's tools and routines. Unlike plain staff augmentation, the provider often stays accountable for the quality of the work.

**Technical debt**: The future cost created by shortcuts, outdated dependencies or design choices that no longer fit. Like financial debt it accrues interest: each change takes longer until the debt is paid down through refactoring or upgrades.

**Technical due diligence**: A review of a product's code, infrastructure, security and engineering team, usually before an investment or acquisition. It tells the buyer what they are really getting, what risks come with it and what it would take to fix them.

**WebSocket**: A protocol that keeps a two-way connection open between a client and a server. The server can push updates the moment they happen, which makes it a common choice for live dashboards, chat and notifications.

**Zero-downtime deployment**: Releasing a new version without making the service unavailable to users. Common techniques include rolling updates, blue-green deployments and backward-compatible database changes.

**n8n**: A workflow automation tool that connects business software and chains actions from a trigger, such as a form or an email. It runs as a cloud service or on your own servers.

**Indexing**: The step in which a search engine analyzes a page it has crawled and stores it in its index. Only indexed pages can appear in search results.

**XML sitemap**: A file listing the pages of a website that its owner wants search engines to find and index, and optionally the date each one last changed.

**E-invoicing**: Invoices exchanged between businesses as structured data through approved platforms, rather than as plain PDFs or on paper. In France, it is being phased in from 1 September 2026.