AI engineer · Bridgetown, Barbados

AI that people actually use.

I design, build and run AI assistants and automation for organisations that can't afford a wrong answer. From the first scoping call to the system in production, and after.

Ask my assistant
Live on this site
What has Raymond shipped that's in production?
ELIA, the citizen assistant for the Barbados Licensing Authority. It has been live on bla.gov.bb since October 2026 and answers licensing questions for the public. He also builds Let's Learn, a study app, and me.
bla.gov.bb
Government assistant, live since 26 Oct 2026
Amini · Protexxa
DevSecOps, cloud and AI engineering today
AWS · GCP
Where the systems run
Bridgetown
Barbados, where I live and work

In production.

Each one is running now. Click through and use it.

Government · Web

ELIA

The Electronic Licensing Information Assistant for the Barbados Licensing Authority. Citizens ask about licences, renewals and fees on bla.gov.bb and get answers drawn from the Authority's own material. Architected and built under Amini Corp.

Desktop app

Let's Learn

A study app that turns your sources into grounded chat, quizzes, glossaries and concept maps, then schedules review so it sticks.

TauriPythonMCP
Try the demo ↗

Web · This site

Portfolio assistant

The assistant at the top of this page. It answers questions about my work from my own records and tells me when someone wants to hire.

AWS LambdaBedrock

What I build.

The whole path: scoping, the build, the cloud it runs on, and keeping it right once people use it.

AI assistants

Grounded in your documents. They cite where an answer came from and say "I don't know" instead of guessing.

RAG, Bedrock, Claude

Automation

The repeat work your team does by hand, moved into workflows that run on their own and tell you when they fail.

n8n, Python, APIs

Cloud and security

Infrastructure built least-privilege from day one, with costs stated per component before anything is switched on.

AWS, GCP, DevSecOps

Scoping

A straight answer on whether AI fits your problem at all, and what the smallest useful first release looks like.

Discovery, MVP plans

How it goes.

  1. 01

    Scope

    One call. You describe the problem. I reply with a plan and a cost, or an honest "not a fit".

  2. 02

    Build

    Short rounds. Each one ends with something working that you can try, not a status report.

  3. 03

    Run

    Monitoring, fixes and answer-quality checks after launch, so it stays right when real people use it.

Essay · 18 min read

Working With AI Without Fooling Yourself

Why AI output that looks right is the main risk, and the checks I use to catch it before it ships.

Read the essay
"If it can give a wrong answer to a citizen, it needs a check that catches it. I build that check first."

Raymond Daniel. DevSecOps engineer and technical programme manager at Amini. Founder of Daniel Enterprises.

More about me

Have a problem AI might solve?

Tell me what you need. I reply within 2 business days with next steps, or a straight "not a fit".

Opens your email app with this filled in. Or email furion128@gmail.com directly.