Tech Glossary

Opposite of 'waterfall' (build everything upfront, deliver at the end). Agile = ship something small, learn, ship more. Scrum is one form of agile.

AI is a broad term covering everything from chess engines to ChatGPT. It includes machine learning, deep learning, and rule-based systems.

Entry-level AI cert from Microsoft. £125, 45 minutes. Covers AI concepts, Computer Vision, NLP, Generative AI, and Responsible AI.

A recipe is an algorithm. So is a search function. The 'algorithm' that decides your TikTok feed is just a complex set of rules.

An API is a set of rules that lets two applications communicate. When you check the weather on your phone, your app calls a weather company's API to get the data. APIs are the backbone of modern software.

Example: The Twitter API lets developers post tweets from their own apps.

Common way SaaS companies measure size. $10M ARR is the threshold investors call 'real'.

If one AZ has a power failure, the others keep running. Modern apps run across multiple AZs for 99.99%+ uptime.

AWS offers hundreds of services for computing, storage, databases, AI, and more. About 31% of all cloud workloads run on AWS. Common entry certification: AWS Cloud Practitioner (CLF-C02).

$100, 90 minutes. Covers cloud concepts, AWS services, security, billing. Great first step into AWS.

Entry-level Azure cert. £125, 45 minutes, multiple choice. Most popular cloud cert for beginners. No technical experience needed.

Azure is Microsoft's answer to AWS. About 25% of cloud workloads run on Azure. Particularly strong in enterprise and government. Common entry cert: AZ-900.

Examples: Salesforce, Slack, AWS. Longer sales cycles than B2C, bigger deal sizes. Most tech jobs are B2B.

Examples: Netflix, Spotify, Amazon retail. Shorter sales cycles, smaller deal sizes, but huge volume.

Handles logic, database access, authentication. Common languages: Python, Node.js, Java, Go, Ruby.

Rule of 3-2-1: 3 copies of data, on 2 different media, with 1 offsite. Most ransomware attacks succeed because the victim has no recent backups.

Measured in Mbps (megabits per second) or Gbps. Higher bandwidth = faster downloads.

Pay per query (per terabyte scanned). Can handle petabytes. Popular for analytics, especially when paired with Google Ads or Google Analytics data.

Term coined in the 1940s when actual moths got stuck in early computers. Every software has bugs. Debugging = finding and fixing them.

Your browser caches images. Your CPU has a cache. CDNs cache content. Caching makes everything faster but introduces 'stale data' problems.

When someone in Sydney loads your site hosted in London, a CDN serves them a cached copy from a server in Sydney. Faster for users, less load for you. Examples: Cloudflare, Akamai, AWS CloudFront.

The product that brought AI into the mainstream. Reached 100 million users in 2 months — the fastest-growing app in history.

Push code → tests run automatically → if pass, code deploys automatically. Tools: GitHub Actions, Jenkins, GitLab CI. Foundation of modern DevOps.

Your browser is a client. Your phone app is a client. The pattern of 'client requests, server responds' is called client-server architecture.

Examples: face recognition, self-driving cars, medical imaging diagnosis, OCR. Powered by deep learning.

Containers are like lightweight virtual machines. They start in seconds and run the same way on any computer. Docker is the most popular container tool.

Cookies remember you are logged in, your language preference, items in your cart. Some are first-party (set by the site you visit), some are third-party (used for ads, increasingly restricted).

Uses SQL, Excel, and BI tools to build dashboards and reports. Lower entry barrier than data scientist; very high demand.

Uses tools like SQL, Python, Airflow, Spark, dbt. Builds the plumbing so analysts and scientists can do their job. Usually highest paid of the data roles.

Cheaper than a data warehouse. You can put any data in (logs, images, JSON, CSVs) and figure out the schema later. Examples: AWS S3, Azure Data Lake Storage.

Uses statistics, machine learning, and programming (usually Python). Typically has a stronger maths and modelling focus than a data analyst.

Examples: Snowflake, BigQuery, Redshift, Synapse. Optimised for reading huge amounts of data quickly. Used by business analysts and data scientists.

Inspired by how brains work. Deep learning powers image recognition, voice recognition, and modern AI like ChatGPT. Requires lots of data and computing power.

Focus on automation, monitoring, fast feedback loops. Tools: Docker, Kubernetes, Terraform, Ansible, Jenkins. DevOps Engineer is a popular high-paid role.

Like a phone book for the internet. When you type a website, your computer asks DNS for its IP address. Cloudflare, Route 53, Google DNS are common providers.

Docker packages an application with all its dependencies into a single 'image'. The image can then run anywhere Docker is installed. Foundation of modern cloud deployments.

EC2 lets you rent computers (virtual machines) by the hour. You choose the size, OS, and region. The most-used compute service on AWS.

Words with similar meaning have similar embeddings. Embeddings let computers do math on language. They are the foundation of search, recommendations, and RAG.

Encryption at rest (data on disk) + encryption in transit (data over the network). HTTPS uses encryption in transit. Modern standard: AES-256.

Extract data from databases/APIs, Transform it (clean, reshape, enrich), Load it into the warehouse. Often automated with tools like Airflow, Fivetran, dbt.

Start with a general LLM (e.g. GPT-4). Train it further on your company's data so it knows your products, your tone, your domain. More expensive than prompting; more accurate for specialised tasks.

Blocks bad traffic, allows good traffic. Hardware (e.g. Cisco) or software (e.g. Windows Firewall). Essential security layer.

Built with HTML, CSS, and JavaScript (or frameworks like React, Vue, Angular). Frontend developers make sure things look good and work smoothly.

Common in small teams and startups. Bigger companies usually specialise developers into frontend or backend.

Smaller than AWS and Azure (about 11% market share) but excellent for big data, machine learning, and analytics. Common entry cert: Cloud Digital Leader.

In effect since 2018. Companies must get consent before collecting data, allow users to delete their data, and report breaches within 72 hours. Fines up to 4% of global revenue. Applies to anyone processing data of EU/UK residents, regardless of company location.

Different from earlier AI which mainly classified or predicted. Generative AI examples: ChatGPT (text), Midjourney (images), Sora (video), GitHub Copilot (code).

Tracks changes to code over time. Lets teams work on the same project without overwriting each other. Created by Linus Torvalds (the Linux guy) in 2005.

Where most open source code lives. Also a social network for developers, an issue tracker, and a CI/CD platform. Essential to know for any dev role.

$49. 10 hours. Teaches prompt engineering and how to use AI tools at work. Perfect for non-technical professionals.

LLMs can invent fake citations, fake quotes, fake URLs. They sound confident because the language is fluent. Always verify AI output for facts.

Every website you visit uses HTTP or HTTPS. HTTPS shows a padlock and protects your data from eavesdroppers. Required for SEO and trust now.

You get a virtual machine; you install your own operating system and software. Examples: AWS EC2, Azure Virtual Machines. Most control, most responsibility.

Tools: Terraform, AWS CloudFormation, Pulumi, Azure Bicep. Lets you version, review, and roll back infrastructure changes.

Two types: IPv4 (e.g. 192.168.1.1) and IPv6 (longer, hexadecimal). Every internet-connected device has one.

Human-readable, easy to parse. Looks like: { 'name': 'Alice', 'age': 30 }. Originally from JavaScript but now used everywhere.

Examples: monthly active users, conversion rate, customer satisfaction score. Every team has 2 to 5 KPIs they track.

When you have lots of containers, you need something to orchestrate them: start them, stop them, restart them when they fail, scale them up and down. Kubernetes (K8s) does all of this. Originally built at Google. Now the industry standard.

Upload a function, and AWS runs it when triggered (e.g. by an API call). Pay only for the milliseconds your code runs. Great for small, event-driven tasks.

Example: Resize images automatically when uploaded to S3.

Measured in milliseconds (ms). Below 100ms feels instant. Above 500ms feels slow. Important for gaming, video calls, real-time apps.

Examples: GPT-4, Claude, Gemini, Llama. LLMs power ChatGPT and similar chatbots. They predict the next word based on context, billions of times, to produce coherent text.

If one server gets all the requests, it crashes. A load balancer distributes traffic evenly. Examples: AWS ELB, Nginx, HAProxy.

Instead of writing rules, you show the computer many examples and let it figure out patterns. ML powers everything from spam filters to fraud detection to recommendation systems.

Something you know (password) + something you have (phone) + something you are (fingerprint). Reduces account takeover by 99%+. Should be on every account that supports it.

Each service does one thing (auth, payments, search). They talk via APIs. Easier to scale, harder to debug. Opposite of 'monolith' (one big app).

Key metric. $100k MRR = $1.2M ARR (annual). Investors love MRR because it is predictable.

Build the MVP first. Get feedback. Improve. Don't build features nobody wants. Popularised by Eric Ries in 'The Lean Startup'.

Made of layers of interconnected 'neurons' (just math). Each neuron does a small calculation. Together they can recognise images, translate languages, generate text.

Covers translation, summarisation, sentiment analysis, chatbots, voice assistants. Modern NLP is dominated by LLMs.

Better for flexible, unstructured data. Examples: MongoDB (documents), Redis (key-value), Cassandra (wide-column), Neo4j (graphs).

Examples: Linux, Python, React, PostgreSQL. Most modern tech is built on open source. Contributing to open source projects helps your career.

Examples: Linux, Python, PostgreSQL, React. Built and maintained by communities. The foundation of nearly all modern tech.

Examples: Heroku, Azure App Service, AWS Elastic Beanstalk. Developers just upload code; the platform handles the rest. Faster than IaaS, less control than SaaS.

Companies hire ethical hackers to try to break in. Anything they find gets fixed before real attackers find it.

Still the #1 cause of breaches. 'Spear phishing' targets specific people. 'Whaling' targets executives.

£125. Tests data preparation, modelling, visualisation, and Power BI Service deployment. High demand in UK and EU.

Connects to data sources, transforms data with Power Query, builds reports and dashboards. Most popular BI tool in enterprise. Cert: PL-300.

Sits between engineering, design, and business. Decides what to build, why, and when. High-paid role. Strong demand for PMs who can work with technical teams.

Same AI, different prompts → different answers. Good prompts are clear, specific, give context, and include examples. A whole job (Prompt Engineer) now exists for this.

Example: 'Write a cover letter' vs 'Write a cover letter for a Cloud Engineer role at a small London startup, emphasising my AWS skills and curiosity'.

Instead of relying only on what it learned during training, the LLM looks up info from your documents in real time. Reduces hallucination. Powers most enterprise AI chatbots.

AWS has 30+ regions worldwide. Picking a region close to your users reduces latency. Each region usually has multiple Availability Zones for redundancy.

Uses standard HTTP methods (GET, POST, PUT, DELETE) and JSON. Stateless: each request stands alone. Almost every modern web API is REST.

(Gain - Cost) / Cost × 100%. If a £200 cert leads to a £5,000 salary bump, the ROI is huge. Most cloud certs have stunning ROI.

S3 stores files (called 'objects') in 'buckets'. Used for backups, images, videos, data lakes. Pay per GB stored and per request. One of the most popular cloud services ever made.

Examples: Gmail, Salesforce, Microsoft 365, Zoom. You pay a subscription and access the software through a web browser. No installation needed.

Example: Google Docs is a SaaS replacement for Microsoft Word.

Recurring monthly or yearly revenue. Examples: Salesforce, Slack, Notion, Stripe. Most modern startups are SaaS. Common metrics: MRR (Monthly Recurring Revenue), CAC (Customer Acquisition Cost).

£125. Covers Zero Trust, Entra ID, Defender, Sentinel, Purview. Great first security cert.

Sprints are usually 2 weeks. Team plans the sprint, builds the work, demos it, and reflects. Most common 'agile' method. Roles: Product Owner, Scrum Master, Developers.

Web servers serve websites. Database servers store data. Game servers run multiplayer games. Today most servers are virtual (running on cloud).

Misleading name: there ARE servers, but the cloud provider runs them. You only write functions. Examples: AWS Lambda, Azure Functions, Google Cloud Functions. Pay per execution.

Examples: Microsoft Sentinel, Splunk, IBM QRadar. Detects threats by spotting patterns in logs from firewalls, servers, applications, etc.

Most cloud SLAs are 99.9% (about 9 hours of downtime per year) or 99.99% (52 minutes per year). If they miss the SLA, you usually get a refund.

Almost every data role requires SQL. Used to filter, group, and join data across tables. The most important data skill you can have.

Example: SELECT name FROM customers WHERE city = 'London';

Strong in interactive dashboards. Now owned by Salesforce. Particularly popular in US tech and healthcare.

Roughly 4 characters or 0.75 words. 'Hello' is 1 token; 'Hello, world!' is about 4 tokens. LLMs are limited by 'context window' (max tokens per request) and you pay per token.

A VPC is like having your own isolated piece of the internet on AWS or Azure. You control which servers can talk to each other and what is exposed to the public.

Used to: access work systems from home, hide your real IP from websites, bypass geo-restrictions. Common providers: NordVPN, ExpressVPN, OpenVPN.

Old model: trust everyone inside the company network. New model: verify every request, no matter where it comes from. Foundation of modern security. Three principles: verify explicitly, use least privilege, assume breach.