Senior Full-Stack Engineer, and the AI work after it
20+ years building robust web systems. These days I work with AI: LLM integrations, agentic workflows, and all the awkward production work between models and the people using them.
Good design happens before the first line of code.
TypeScript · React · Next.js · PHP · Laravel · PostgreSQL · Vector DBs · LLM APIs · RAG · Agents · Docker · Kubernetes · Terraform
Paul Radford is a senior full-stack engineer who turns ideas into working software. For more than twenty years he has done the unglamorous, essential work of understanding what people actually need and building it, drawing on a career that spans not just the full web stack but the businesses around it. That instinct for translation runs through everything from early custom CRM builds to enterprise platforms handling hundreds of millions in transactions.
More recently his focus has moved to AI. He has led LLM adoption across his current company, building custom Model Context Protocol integrations into its software so that customer-service, product, engineering, and DevOps teams work with less friction. It is the practical, often awkward production work between a model and the people relying on it.
That AI work sits on a deep technical foundation: rebuilt underperforming legacy systems into scalable platforms, modernised codebases serving hundreds of thousands of users, and bridged the gap between security, operations, and development teams on the infrastructure that ships a product.
Bridging security, operations, and development, and leading LLM and Model Context Protocol adoption across the product.
Rebuilt CRM and external-API performance, and built the platform that supported a major merger across VoIP, NBN, and mobile services.
Rebuilt underperforming legacy applications into scalable platforms, and led a shared-hosting-to-AWS migration backed by Elasticsearch and Redis.
Hands-on PHP, LAMP, and front-end development across agencies and small businesses, including custom CRM and web-systems builds.
Firewall configuration is unforgiving: rules are dense text, ordering bugs are easy to introduce and hard to spot, and one mistake can expose or black-hole a network. RuleForge turns that text into a visual canvas — drag out the network topology, compose rules against it, and simulate how a packet traverses the rule set before anything ships. Complex policies become legible at a glance, and mistakes surface in the simulator instead of in production.
Explore RuleForgeBun makes JavaScript backends far faster to start and run, but the ecosystem still lacks the batteries-included structure that makes a framework like Laravel productive for real applications. Bunary brings that structure to Bun — cohesive and convention-driven, built around Bun's strengths and optimised for developer velocity and minimal cold starts rather than retrofitting a Node-era framework.
View the repoPractical front-end knowledge — the small CSS and HTML techniques that actually ship features — is scattered across blog posts, threads, and gists, and goes stale quickly. template.tips collects them into one maintained library: a React front end backed by a headless Payload CMS, curated and updated weekly so the snippets stay current and copy-paste ready.
Explore template.tipsAI tooling moves fast, and the genuinely useful prompts, tips, and workflows are buried in noise that is outdated within weeks. aitool.tips curates them into a maintained library — the same React and headless Payload CMS stack as template.tips — focused on practical, current techniques for developers working with AI tools day to day.
Explore aitool.tips