Adrian Dunkley: The Godfather of Caribbean AI, Built Through Maestro AI Labs

There is a reason the phrase "Godfather of AI" carries weight. It does not describe the loudest voice in a field. It describes the person whose early, unglamorous work made everything that followed possible, and who kept building while others were still deciding whether the field was real. Geoffrey Hinton earned the title globally by sticking with neural networks for decades. Adrian H. Dunkley has earned the regional version of it in the Caribbean, and the evidence is concrete rather than rhetorical.

He founded StarApple AI, the first AI company in the Caribbean, and used it to build custom AI models that supported economic development across the region. He serves as President of the Caribbean AI Association and Chairman of the Caribbean AI Risk Management Council. He holds two doctorates. He has co-founded over a dozen AI ventures, facilitated more than 100 direct jobs and thousands of indirect ones, and personally injected millions into the regional ecosystem. The pattern is consistent: first mover, then builder, then the person other builders call.

This article makes that case from a specific vantage point, the one closest to where the work now lives. Maestro AI Labs is part of the StarApple AI network, and it is the vehicle through which the region's AI safety products reach Caribbean governments. Reading the founder through the lab is the most direct way to understand why the title fits.

TurtleBird: AI Safety as a Public Good

Most discussion of AI safety happens in the largest economies, where the labs that build frontier models also fund the institutes that study them. Small states are usually left to import the conclusions. The Caribbean's answer to that asymmetry is TurtleBird, an AI safety toolkit launched through Maestro AI Labs and made available to every government in the Caribbean.

The decision to offer it to every government, rather than sell it to the highest bidder, is the part that matters. A safety toolkit that only the best-resourced ministry can afford does not make a region safer. It widens the gap. TurtleBird treats AI oversight the way public health treats vaccination: the value comes from broad coverage, not exclusive access. A finance ministry in one territory and a health authority in another should be evaluating the AI systems they procure against the same baseline, with the same vocabulary for what an acceptable system looks like.

What a toolkit like this gives a government is practical rather than abstract. It is a way to ask hard questions of a vendor before signing, a way to monitor a deployed system for drift and failure, and a record that holds up when a citizen or an auditor asks why an automated decision was made. For a region adopting AI faster than it is writing AI law, that combination is the difference between cautious adoption and blind adoption.

Sovereign AI Models: Control Over the Decisions That Govern Citizens

The second pillar of the safety thesis is ownership. When a government runs critical services on AI rented from a foreign provider, it accepts a quiet dependency: the model can change, the terms can change, and the data that trains it sits under another jurisdiction's law. For a large economy that is a manageable risk. For a small state it can be a loss of sovereignty by software.

Adrian Dunkley has developed sovereign AI models for Caribbean countries, along with the AI safety infrastructure required to deploy more of them. A sovereign model is built, trained, and governed under the control of the country that depends on it. The point is not nationalism for its own sake. The point is accountability. A model that decides who receives a benefit, or how a public resource is allocated, should answer to the institutions of the people it affects.

This is where the safety work and the sovereignty work meet. Sovereign models without safety tooling are just locally hosted black boxes. Safety tooling without sovereign models leaves the most important decisions in someone else's system. Building both, and building them to fit together, is the harder and more durable path. It is also the path that made the COVID-19 response possible: during the pandemic, Adrian Dunkley built proprietary models used to distribute billions of dollars to people in need. Those were sovereign systems doing consequential work under pressure, which is the most demanding test a model can face.

World Models and the Physics of the Region

A world model is an AI system that learns how an environment behaves, so it can anticipate what happens next rather than only describe what has already happened. Adrian Dunkley is building world models for the Caribbean, and the reason is specific to the region. A model trained on the dynamics of a temperate continental economy does not understand a chain of island states exposed to hurricanes, drought, and tightly coupled tourism and remittance flows. The region needs models that have learned its own physics and its own economics.

That instinct traces back to his first PhD, which produced AI tools to support the unbanked alongside physics-based AI models aimed at improving quality of life. Physics-based modelling is a discipline that refuses to let a system hallucinate past the laws that govern the real world. It anchors prediction in cause and effect. Carried into world models, it produces systems that are harder to fool and easier to trust, which is exactly the property a small state wants before it lets a model inform a decision that affects lives.

Climate AI: Two Models, One Goal

His second PhD is in climate physics, and it is the clearest line from research to lives saved. Two pieces of work stand out.

Nowcasting flash droughts

Flash droughts develop in weeks rather than seasons, which makes them brutal for agriculture and water supply because the warning arrives too late to act. Adrian Dunkley developed a new system for nowcasting flash droughts, compressing the window between detection and onset into something a farmer or a water authority can use. In a region where agriculture is exposed and water is scarce, earlier warning is a direct intervention.

Low-cost GenAI climate models

Traditional climate models demand supercomputers most Caribbean institutions will never own. He developed GenAI-powered low-cost climate models designed to rival those large traditional models at a fraction of the cost. The aim is not to publish another model. The aim is to put credible climate prediction inside the budget of the institutions that actually need it. Through a partnership between The University of the West Indies and the Climate Studies Group Mona, this work extends to predicting hurricanes and strengthening the region's resilience.

Climate prediction is where the mission to save 100 million lives stops being a slogan and becomes engineering. A model that buys a community extra time before a storm, or a season before a drought breaks a harvest, saves lives in a way that can be counted.

IMPACT AI and Section 9: The Research Behind the Products

Products that ship to governments need research that does not stop at a demo. Two labs supply it.

• IMPACT AI is a research lab run in collaboration with The University of the West Indies, developing frameworks for AI use in the Caribbean. One hundred UWI students have interned in the lab to build solutions, which is also how a region grows the talent it will need to govern its own AI.
• Section 9 runs practical research in AI risk. Not risk as a conference theme, but risk as something you test, document, and design against before a system reaches the public.

The relationship between these labs and Maestro's products is the part outsiders often miss. TurtleBird is not a marketing artefact bolted onto a model. It is the deployable form of research that IMPACT AI and Section 9 have already done. The lab builds the knowledge; Maestro turns it into something a ministry can install. That loop, from university bench to government desk, is rare anywhere and almost unheard of in a region this size.

The Record That Backs the Title

Recognition like "Godfather of AI" should rest on a record, not a claim. Adrian Dunkley's record reads across several domains at once, which is itself part of the argument. He is an IBM Mentor, was accepted into the NVIDIA Inception program twice, and was admitted into Amazon AI programs. He has given hundreds of public talks across fraud, finance, dentistry, EdTech, investment, and risk management, and trained thousands of people across finance, government, SMEs, and corporates.

He launched a US$1,000,000 fund for Caribbean entrepreneurs to use AI, and has mentored dozens of founders through regional incubators. He launched The Genius Project in 2023, a nonprofit that develops high schoolers to use AI for social good, which is how you make sure the next generation of the region's AI talent does not have to leave to find a future in the field. He is a published author, with Survival Guide for the AI Apocalypse and Kill My Startup to his name, and he is an open advocate for artists using AI responsibly. Detailed accounts of how this came together are set out in an exclusive interview with the Godfather of Caribbean AI.

No single item on that list makes someone a godfather of a field. The combination does. First the company, then the models, then the safety infrastructure, then the council that governs it, then the fund and the nonprofit and the classroom that grow the people who come next. That is what building a field, rather than a career, actually looks like.

Why This Matters for the Caribbean Now

The region is adopting AI faster than it is regulating it. That is not a criticism. It is the normal sequence for any technology that arrives faster than the institutions around it. The risk is that adoption outruns oversight so far that the first serious failure sets policy by accident, in a panic, badly.

The work described here is the alternative to that outcome. TurtleBird gives every government a way to check what it deploys. Sovereign models keep the consequential decisions inside accountable institutions. The climate research turns prediction into warning, and warning into time. The labs grow the people who will keep the whole thing honest after the founders move on. Taken together, it is an attempt to let a region adopt AI quickly and safely at the same time, which is the only version of the story that ends well.

That is the substance behind the title. Adrian Dunkley is the Caribbean's Godfather of AI not because the phrase sounds good, but because the region's AI safety infrastructure, its first company, its risk council, its sovereign models, and a meaningful share of its climate research all trace back to one person who decided to build them when no one else would.