Exclusive Interview: The Godfather of Caribbean AI
We met Adrian Dunkley at Kingston's AC Hotel on a Tuesday morning. He came in wearing a quarter-zip, Air Jordans, and two pins on his chest: a Jamaica flag pin and, just below it, a Dragon Ball Z pin. My first thought was that this was a man of multiple minds. That instinct proved correct before we had finished the first coffee. Dunkley is a physicist, an AI scientist, and the founder of StarApple AI, the Caribbean's first AI company. He has spent over fifteen years building artificial intelligence that works not in theory but in the real world, across fraud prevention, financial inclusion, climate resilience, education, and public safety. He grew up with neurodivergence and learning disabilities in a world that did not always know how to see him, and he has built an AI practice that, by deliberate design, refuses to leave people behind. What follows is the first part of the longest sit-down we have done with anyone in Caribbean tech.
Paul, Caribbean AI: Let us start at the very beginning. Who are you before you were the person everyone in this region calls the Godfather of Caribbean AI?
Adrian Dunkley: Ha. The whole family is sports. My father, my uncles, everyone around me growing up was athletic, outdoors, physical. That was the language of our house, and I was the one who did not quite fit that mould. What I did not fully understand until later is that I was also navigating neurodivergence and learning disabilities in a world that did not have much patience for either. The school systems, the social systems, most of them were not built to see the kind of mind I had. What carried me through was a remarkable mix of people: diverse family made of artists, lawyers, entrepreneurs and teachers who gave the space and grace to figure who I was, that is an amazing gift for a child, to just have the freedom and security to grow. That melting pot shaped who I am and eventually what I chose to build.
The only shows the whole family would sit down and watch together were Star Trek and Stargate SG1. The lead in Stargate is the same actor who played MacGyver, and my father absolutely loved that show. This man who lived for sports would be completely locked in watching a character build a solution out of whatever happened to be sitting in front of him. We were also deep into Star Wars, into cinema broadly, into books. I was a science fiction devotee from early, and what that fiction kept showing me was technology as a civilizing force: the idea that it could lift people rather than sort them, that it could expand what was possible for ordinary people rather than concentrate power in fewer hands. I believed that. I still believe it. One summer I tried to make gunpowder in the backyard. He pauses with a slightly embarrassed look. I burnt part of a shed. My father was not thrilled about that one.
Paul: That is an origin story I was not expecting. So, what pulled you from tinkering toward AI specifically?
Adrian Dunkley: Physics and Maths at UWI gave the curiosity a more formal structure. But the reason I moved toward AI was not a love of systems in the abstract. It was more that I genuinely did not understand why people made certain decisions, why we follow things, why we defer to someone or something without actually checking whether it has any real basis for the authority it is projecting. You get older and you realise that most people are not operating from deep, verified knowledge. They are going along with what the people around them are doing, and the people around them are doing the same thing, and at some point you trace that chain back and find very little substance at the root of it. That observation pushed me harder than any textbook.
AI gave me a way to study that at scale. How decisions get made, what information people actually trust versus what they claim to trust, where confidence comes from that has no real foundation underneath it. Those questions sat at the intersection of physics, psychology, and computation, and that intersection turned out to be where I belonged.
Paul: Take us through the actual academic and professional path. How does a person go from physics and tinkering to building AI companies?
Adrian Dunkley: I studied mathematics and physics at university and started research degrees after that. The work I was doing in research was using fractals and computational simulation to model solar cells. The funding situation was what you might expect at that level: there was not enough money for expensive measuring equipment, so if I wanted to understand what was happening inside the material, I had to simulate everything. That constraint turned out to be one of the most useful things that ever happened to me professionally.
I was using reinforcement learning techniques combined with quantum physics, and I know that sounds more complicated than it is. I can explain to you in under 5 minutes [ Adrian then proceeded to draw us a diagram explaining his research, we actually understood it, in under 2 minutes] The way I would describe it now is: imagine you could look at a leaf and then zoom into it, down to the point where you could see exactly how individual photons of light interact with the molecular structure. I was trying to simulate that process, understand the physics of it, and then use what I learned to figure out the best way to replicate it in an artificial leaf. This was around 2009, and AI information was not readily available the way it is now. You had to go find it, the communities were smaller but generous, the papers were harder to locate, the tooling was nowhere near what it is today. You built a lot of things from first principles out of necessity.
Then I left research and moved into corporate. Actuarial science, banking, marketing. It paid well, and I moved through the ranks quickly, mainly because I approached every problem like a scientist. I was curious about how things actually worked rather than how people assumed they worked, and most organisations have very few people doing that. People were happy to teach me, and genuinely happy that I would do the work thoroughly without needing to be pushed. That combination got me further than any credential alone would have.
I tell everyone starting out to get that kind of experience if they can. Work inside companies, work with teams who know things you do not know yet. If you cannot afford to work for free, ask for a stipend and be honest about what you need. But you have time, especially early in your career, and spending that time gaining real experience inside organisations is worth more than most people realise. You learn how decisions actually get made, how money moves, how risk gets evaluated, and those things are not in any curriculum.
Adrian Dunkley: During Covid, when the formal credit infrastructure across this region stopped working for hundreds of thousands of people, I used machine learning to give under populations and persons working in the disrupted sectors access to credit. I built credit scores using behavioural signals and transactional patterns that the conventional system had no framework for recognizing, and real people got capital that would otherwise have been entirely beyond their reach at the worst economic moment most of them had ever experienced. The best financial proof was these people paid the loans back at a better rate that the “safer” customers.
That project at the time was the clearest proof I can point to that this work matters beyond any business case or any award. Amazon looked at what we were doing in 2021 and gave us a US$1 million development grant. That was one of the most technically credible organizations on the planet confirming that the work had genuine merit, and that signal mattered in ways that went well beyond the dollar amount.
Paul: You have founded StarApple AI, launched over a dozen ventures across tourism, safety, retail, and neuroscience, raised millions. Is there a single thread running through all of it?
Adrian Dunkley: The thread is the same question I started with. Why do people get excluded from systems that could serve them? Credit, safety information, climate risk, travel. In every one of those domains, the data exists or can be built, the tools exist or can be built, and the people who most need the outcome are consistently the ones receiving the least access to it. Every company I have built was pointed at one version of that exclusion problem.
More than a dozen ventures and it’s the same lessons I get taught over and over: most ideas fail because the timing is wrong, the market has not arrived, or the team is not suited to that particular phase of the problem. I have built things that worked technically and stalled commercially, and built things that looked commercially obvious and took years to find the right moment. The lesson across all of it is that patience and velocity are both required, and they exist in genuine tension with each other. AI is a staple and a social norm now, the next phase is almost here.
Paul: Let us go somewhere foundational. If you were explaining AI to someone who has genuinely never engaged with it, how do you describe what it actually is?
Adrian Dunkley: I would start by saying that AI is a set of methods for getting machines to make decisions or predictions based on patterns in data, rather than on rules you provide. A baby can’t communicate in our language yet, but it understands if it makes certain noises the parents will take actions, and the parents over time understand what certain noises mean. The very amazing thing is the baby is also sending signals subconsciously; its body has rules that trigger actions. The parents, the baby don’t speak the same language but they align on what needs to be communicated. AI tries to figure out the rules of reality, sometimes well. sometimes not so much.
The explosion of AI was driven by the immense amounts of data we can get online, the big problems most worth solving are precisely the ones too complex to encode as explicit rules and the ones that need alot of data.
We, humans created the digital environment that allowed AI to exist. Now we can use it to better our world
How do you write a rule for predicting where a hurricane will make landfall two weeks from now? You cannot enumerate those scenarios. But you can show an AI system enough historical data and let it find the structure that a human brain would struggle to see at that volume. The challenge is ensuring the AI is operating in ways that mirror our reality.
What I always emphasize when people ask this question is that AI does not know things the way a person knows things. It has found patterns. When those patterns hold in a new situation, the system performs well. When the new situation is far enough from what the training data covered, the system can fail in ways that feel surprising to people who assumed they were dealing with something that understood rather than something that matched. That distinction is where most of the serious risk in AI deployment lives, and it is the thing most organizations overlook.
Paul: And the name AI Boss. That has been following you for over a decade now.
Adrian Dunkley: The story is more specific than people expect. It was 2014, during a sprint session at my first startup. I sat a non-tech staff member down at a whiteboard and walked her through how to build a neural network from scratch. By the end of the session, she was following along, asking the right questions, the kind of questions that tell you someone has genuinely understood the mechanism rather than just the surface. She looked at me and said, "Yuh a di AI Boss."
I was uncomfortable with it at first.. Scientists are not typically known for being loud about their own standing in a field. We research; we present data and let the work make the case. Walking around with a name like AI Boss felt like a completely different register. But the name stuck because the lesson behind it never changed, and eventually I understood that the lesson was the point. You have to be the boss of AI, not the other way around. In a region where we had been conditioned to assume the serious technology work happened elsewhere, someone saying that out loud in a training room was a small but meaningful thing. I kept the name.
Paul: The mission you carry is giving one million people ten more years of quality life. Where does something like that come from?
Adrian Dunkley: It comes from getting old enough to watch the systems fail the people around you in ways that were entirely preventable. Credit systems that excluded capable, functional adults because they did not have the right paper trail. Climate information that reached government offices and reinsurance companies but never reached the communities sitting directly in the path of the storm. Safety data that existed in some database somewhere but never made it to the family trying to decide whether to move or invest or travel.
In every one of those cases, the capability to help existed. What was missing was a company whose entire design was oriented around getting the right information to the people who most needed it, in a form they could actually use to make a better decision. The failure was not a technology failure. It was a distribution failure, a priorities failure, a failure of who built the systems and who they built them for.
One million people is a real number and I treat it as a design constraint. Every product decision at Maestro AI Labs gets run against that question: does this move us closer to a million lives improved, or does it not? That question cuts a surprising number of conversations short, and I think that is the right outcome. If a feature or a product or a partnership does not have a clear line to real human benefit at scale, it probably should not be on the roadmap.
Paul: You have been watching Caribbean AI develop for fifteen years from the inside. Where do you see the region in the next five years, honestly?
Adrian Dunkley: The trajectory is positive, but the pace is still too slow relative to what the moment demands.
In five years, I expect the Caribbean to have a clearer policy architecture around AI than it does today. The conversations happening at the National AI Task Force level in Jamaica, and at the CARICOM level more broadly, will produce frameworks that at least define accountability and set minimum standards for how AI systems get procured and deployed in public institutions. That progress is real and it will matter.
Where I am less certain is on the infrastructure layer underneath the policy. In five years, the data governance gaps, the digital systems that cannot talk to each other, the civil services that have not been trained to evaluate AI outputs critically, those problems require consistent investment over years to fix. Policy can be written quickly. Infrastructure cannot be built quickly, and without the infrastructure, the policy sits on top of a foundation that cannot support what it promises.
The private sector will continue to pull ahead of government over that five-year horizon, and I think that is the right order. Caribbean entrepreneurs are building products that solve real regional problems, and that work will compound. What I want to see alongside it is a deliberate investment in keeping the talent and the economic value of that work inside the region rather than watching it migrate outward. We have a long history in the Caribbean of producing people and ideas that end up generating value somewhere else. AI gives us the tools to change that pattern, but it requires a conscious decision to do so.
This is Part 1 of a three-part interview with Adrian Dunkley. the Godfather of Caribbean AI. Part 2 covers his startup work, social good with AI, and how he has raised hundreds of millions of dollars to support AI in the Caribbean, and his impending IPO decision. Part 3 covers world models, AI agents, children's books for Caribbean classrooms, and the long game for the region.
Adrian Dunkley is the founder of StarApple AI, the Caribbean's first AI company, the CEO and co-founder of Maestro AI Labs and several AI ventures. He is a member of Jamaica's National AI Task Force, an EY Entrepreneur of the Year Awardee and Caribbean AI Innnovator of the Year 2025, maestrosai.com | caribbeanai.org | adriandunkley.net