Engineering & Delivery
Day-to-day terms you will see in standups, PRs, and SOWs.
MVP Minimum Viable ProductThe smallest version of a product that delivers a learning. The goal is not minimal, it is enough to validate or invalidate a hypothesis. See our MVP development service.
Trunk-Based DevelopmentA branching strategy where every engineer integrates work into one long-lived branch (main) at least once a day. Feature branches live less than 48 hours. Reduces merge pain and drift.
Definition of DoneA team-agreed checklist a piece of work must satisfy before it is marked complete: tests, docs, security scans, accessibility checks, observability. See our methodology page.
CI/CD Continuous Integration / DeliveryAutomation that runs every commit through tests, security scans, and packaging, then promotes good builds to staging and production without human handoff.
Technical DebtThe cost of choosing a faster but worse implementation today, paid back in slower future changes. Manageable; only dangerous when it is invisible. See our blog post.
Feature FlagA runtime switch that lets a feature be merged into main without being visible to users. Enables daily merges without daily releases.
Code ReviewThe practice of having every change inspected by a second engineer before merging. We pair it with an AI reviewer agent. See our methodology.
AI & Machine Learning
Terms you will see whenever an LLM, agent, or model is involved.
LLM Large Language ModelA neural network trained on huge text corpora that predicts the next token. Claude, GPT-4o, Gemini, and Llama 3 are LLMs. Useful when generation, classification, or extraction is needed at scale.
RAG Retrieval-Augmented GenerationAn LLM pattern that retrieves relevant documents from a vector store and inserts them into the prompt at query time. Reduces hallucinations and grounds answers in source material. See RAG vs Fine-tuning.
Fine-tuningContinued training of a pre-trained model on a smaller, task-specific dataset. Useful when style, format, or domain-specific behaviour matters more than raw knowledge.
EmbeddingA high-dimensional vector that represents the meaning of a piece of text, image, or audio. Similar items have similar vectors. The basis of vector search.
Vector DatabaseA database optimised for nearest-neighbour search over embeddings. pgvector, Pinecone, Weaviate, Qdrant, Milvus are common choices.
Agent / Agentic AIAn LLM-powered system that can plan, call tools, and iterate on its own work. Different from a single-shot chat. Examples: Claude Code, Cursor agents, browser-use.
MLOpsThe discipline of running machine learning models reliably in production: versioning, monitoring, drift detection, retraining triggers, rollback. The DevOps of ML.
HallucinationWhen an LLM generates plausible but incorrect output. Mitigated with RAG, citations, smaller scopes, and human review on high-stakes decisions.
TokenA unit of text an LLM processes (roughly 0.75 words). Models have token limits per prompt and per response. Costs are usually billed per million tokens.
VoIP & Telecom
The vocabulary of softswitches, SIP, and contact centres.
SIP Session Initiation ProtocolThe signalling protocol used to start, manage, and end voice or video calls over IP. Think "HTTP for phone calls".
RTP Real-time Transport ProtocolThe media protocol that carries the actual audio/video packets once SIP has set up the call.
FreeSWITCHAn open-source softswitch used to build voice applications, contact centres, and PBXs. Strong in mixed-media and custom IVRs. See our FreeSWITCH developer page.
AsteriskAn older open-source PBX framework that powers FreePBX and many on-prem call systems. See our Asterisk developer page.
Kamailio / OpenSIPSHigh-performance open-source SIP servers used as proxies, registrars, or load balancers in front of softswitches. See Kamailio / OpenSIPS.
SBC Session Border ControllerA specialised SIP proxy that sits at the edge of a VoIP network providing security, NAT traversal, and protocol translation.
WebRTCA browser-native standard for real-time audio, video, and data. The basis of in-browser calling, video conferencing, and live-streaming. See our WebRTC developer page.
SIPpAn open-source load-testing tool for SIP. We use it to validate FreeSWITCH and OpenSIPS deployments under load.
Cloud & DevOps
Infrastructure, deployment, and reliability terms.
IaC Infrastructure as CodeDefining cloud resources in version-controlled code (Terraform, Pulumi, CloudFormation) so infrastructure is repeatable, reviewable, and auditable.
Kubernetes / K8sAn open-source container orchestrator that schedules, scales, and heals containerised workloads. Standard for cloud-native deployments.
MicroservicesAn architecture style where the system is split into independently deployable services. Trade-off: more flexibility, more operational overhead.
SRE Site Reliability EngineeringThe practice of running production systems using software engineering tools and discipline. Owns error budgets, SLOs, and on-call.
ObservabilityThe ability to understand what a system is doing from its outputs: logs, metrics, traces. Different from monitoring, which only catches what you predicted.
Blue/Green DeploymentA deployment pattern where a new version is deployed alongside the old, traffic is switched over once verified, and the old version is kept as instant rollback.
Canary ReleaseReleasing a new version to a small percentage of users first, watching for regressions, then ramping the percentage up. Reduces blast radius.
Data & Analytics
From pipelines and warehouses to causal inference.
ETL / ELT"Extract, Transform, Load" vs "Extract, Load, Transform" — two patterns for moving data into a warehouse. Modern warehouses (Snowflake, BigQuery) favour ELT.
Data Warehouse vs Data Lake vs LakehouseWarehouse = structured analytical store (Snowflake, BigQuery). Lake = raw storage of any format (S3, GCS). Lakehouse = the two patterns combined (Databricks, Iceberg).
dbtA tool that lets analytics engineers transform data in the warehouse using SQL + version control + tests. The Git workflow for analytics.
A/B TestAn experiment that randomly assigns users to two variants to measure causal effect of a change. Done right, it answers "did this work?" with statistical rigour.
Cohort AnalysisGrouping users by a shared characteristic (signup month, plan, geography) to compare behaviour over time.
Compliance & Security
Frameworks and acronyms you will see in security questionnaires.
SOC 2A US compliance framework for service organisations covering security, availability, processing integrity, confidentiality, and privacy. Type 1 = controls exist; Type 2 = controls have operated effectively over a period. See our compliance page.
HIPAAUS healthcare privacy and security law. Requires technical, administrative, and physical safeguards over Protected Health Information (PHI).
GDPREU/UK data protection regulation. Requires lawful basis for processing, data subject rights, breach notification within 72 hours, and DPAs with sub-processors. See our DPA.
PCI DSSPayment Card Industry Data Security Standard. Required for any system that stores, processes, or transmits card data. Scope-reduction (tokenisation) is the cheapest path to compliance.
ISO 27001An international standard for information security management systems (ISMS). Audited annually, recertified every 3 years.
Pen Test Penetration TestAn authorised simulated attack on a system to find exploitable weaknesses. Annual is standard for SaaS; quarterly for regulated workloads.
SAST / DAST / SCAStatic, dynamic, and software-composition analysis — three classes of automated security scans. We run all three in CI (Semgrep, OWASP ZAP, Snyk).
Commercials & Engagement
The contractual and pricing vocabulary of vendor engagements.
NDA Non-Disclosure AgreementA bilateral or unilateral contract preventing one or both parties from disclosing confidential information. Standard before any commercial-in-confidence call.
MSA Master Services AgreementThe umbrella commercial contract between a buyer and a vendor. Defines payment terms, IP, liability, and the rules for any future Statement of Work. See our sample MSA.
SOW Statement of WorkA per-project schedule signed under an MSA. Defines scope, deliverables, milestones, fees, and assumptions. Where the work actually lives.
DPA Data Processing AddendumAn addendum required under GDPR/UK GDPR when one party processes personal data on behalf of another. See our DPA template.
Fixed Price vs Time & MaterialsTwo pricing models. Fixed = vendor takes the scope risk for a fixed fee. T&M = buyer pays for hours worked at agreed rates. See our blog post.
RetainerA recurring fee that buys a guaranteed pool of senior engineering hours per month. Useful when scope shifts week to week.
RFP Request for ProposalA buyer's formal request for vendor responses to a defined need. We publish a free RFP template.
SLA Service Level AgreementA commitment to specific performance targets (response time, uptime, resolution time) with consequences for missing them. See our SLA page.