A Curriculum for Caribbean Women Who Want to Build

Note from the President

March is Women's History Month, and a good moment for an honest accounting. CAIRA's membership crossed seven thousand last week. Women make up thirty four percent of that membership. Women hold three of the eight seats on our Board. Women lead four of our seven working groups. The numbers are not bad for a region whose technology sector has been male-led for most of its history, but they are not the numbers we want either. We have spent more time this quarter on the question of how to change them, and I expect to spend more time on it next quarter.

Kwame Stevens has written our feature this month, and it is a different kind of piece than we have run so far. It is, in plain terms, a curriculum. Kwame has been teaching Caribbean students for twenty years, and he has put together a sequence of lessons that any Caribbean woman, of any age and any starting point, could follow over six months to move from a position of using AI to a position of building with it. We will publish a longer version on our learning platform in April. This issue contains the outline.

Our Charter draft remains open for member comment until the end of the month. The first general public consultation will be held in Bridgetown on March 22. Sign up at the usual address.

Adrian Dunkley Founder and President, Caribbean AI Association

---

Feature

A Curriculum for Caribbean Women Who Want to Build

By Kwame Stevens

I want to do something different in this column. I am not going to argue that more Caribbean women should be building AI products. The argument has been made well and often. I am going to assume you have heard it, and I am going to give you the next thing. The next thing is a path.

By the end of this article, you should know what to learn, in what order, in what time, with what tools, and to what end. If you are a Caribbean woman who has decided to move from being a user of AI to being a builder, this is the plan. If you are reading this on behalf of a daughter, a sister, an employee, or a student, the plan is the same.

A note before I begin. The curriculum I am describing was assembled with three audiences in mind. The graduate of a Caribbean secondary school who is wondering whether to study computer science. The mid-career professional, in any field, who has decided to learn AI in the next year. The returning student, often a woman taking up new work after children have grown. I have taught all three. The path I describe is the same path, with different paces and different anchors. If you are a complete beginner, expect this to take six months at six hours a week. If you are a working professional with technical experience, expect it to take three months at the same intensity. You will not need to leave the Caribbean to complete it. You will not need to spend more than four hundred US dollars in tools and learning materials.

Month one. Understand what you are working with.

The first month is the month of learning to see. Pick one large language model, I would suggest Claude or ChatGPT for this purpose, and use it every day for at least an hour, on a task you actually have to do. Write the assignment. Plan the trip. Answer the email. Summarize the meeting. The goal is to develop intuition, not knowledge. By the end of the month you should know what these tools do well, what they do poorly, what they invent, what they refuse, and what their limits look like in practice.

In parallel, read three things in the first month. The first is the Anthropic prompt engineering guide, which is the clearest plain language introduction to how these tools think. The second is the OpenAI documentation on structured outputs and tool use, which will be the basis of what you build later. The third is the Hugging Face introduction to transformers, which goes one layer deeper than the marketing material and will save you from believing a great deal of internet nonsense.

You should not pay for a course in month one. The free materials are excellent.

Month two. Learn just enough programming.

The next month is the month where most people quit. I am going to ask you not to. Caribbean women are over-represented in the population of people who taught themselves to read English from books they did not initially understand. You have already done the harder version of this exercise. Programming is easier than English.

Learn Python. Specifically, the parts of Python that are useful in building with AI. Variables, lists, dictionaries, functions, error handling, working with JSON, reading and writing files, making HTTP requests, using libraries. You do not need to learn classes, inheritance, or any object-oriented theory in month two. You can come back to those.

The fastest path I have found is the free Python course at Boot.dev, paired with the free Real Python tutorials. Plan ten hours a week for four weeks. Use Claude or ChatGPT as your tutor when you get stuck. Do not copy the answer. Ask the model to explain the error, and then fix the code yourself. This is the most important discipline of the curriculum and the one that separates the builders from the users.

Month three. Build your first AI tool.

Pick a small problem you have. The smaller the better. Mine, when I taught this curriculum to the first cohort in Spanish Town in 2023, was a tool that read my children's school newsletters and pulled out the dates of upcoming events into my calendar. Your problem can be anything. Reading the WhatsApp parent group messages and summarizing the school news. Drafting reply emails in your own voice. Categorizing the receipts in your business expense folder. Generating a weekly menu plan based on what is on sale at the supermarket.

In month three, you build that thing. The architecture is the same regardless of the problem. A small Python script. An API call to a language model. A prompt that does the actual work. A loop that processes inputs and produces outputs. A way to store or display the result.

The technical skills you will pick up in this month are the ones that matter. How to make an API call. How to manage an API key without leaking it. How to write a system prompt. How to handle the cases where the model returns something you did not expect. How to test what you have built. Plan twenty five hours of work over the month. You will hate it for ten hours, get into it for ten more, and finish the last five with the satisfaction of someone who has made a thing that did not exist before.

Month four. Learn the platforms.

By month four you have built one thing. You will want to build more, and you will want to build them in ways that other people can use. This is the month you learn the modern AI development stack.

Specifically, you should learn the basics of three things. First, a deployment platform. Vercel and Replit are both good starting points for Caribbean developers because they require no credit card sophistication and they work well from our region's typical internet conditions. Second, a vector database, for which I would recommend Pinecone or Supabase. Third, a way of orchestrating AI agents that take multiple steps, for which the simplest current option is LangChain or the OpenAI Agents SDK.

You will not master these in a month. You will learn the shapes well enough to know which tool fits which job. That is the goal.

Month five. Build something useful to someone else.

This is the month the curriculum stops being about you and starts being about your community. Pick a person, an organization, or a small business in your network, ideally one that is not technology native. Talk to them about their work. Find a specific repeated task that costs them time, that involves words or documents, and that an AI tool could plausibly handle. Build that tool for them. Deploy it. Watch them use it.

I have run this exercise with thirty Caribbean students and the outcomes have been remarkable. One student built a tool for her aunt's catering business that converts WhatsApp customer enquiries into structured invoices. Another built a tool for her father's law office that reads scanned land registry documents and extracts the relevant clauses. A third built a tool for her secondary school that helps the guidance counsellor match students with scholarship opportunities. None of these were technically sophisticated products. All of them did real work for real people.

The point of month five is to learn the muscle of building for someone else's problem rather than your own. That is the muscle of the professional.

Month six. Choose your direction.

By the end of month six you will know enough to make a choice. The choice is which kind of builder you want to be.

You could become an AI engineer, which means going deeper on the systems and the infrastructure, learning Kubernetes and cloud platforms, learning to train and fine tune models, and ending up in a role at a software company or a large enterprise. The starting salary in this role across the Caribbean, paid in US dollars by remote employers, runs from sixty thousand to one hundred and twenty thousand at the time of this writing.

You could become a product builder, which means going deeper on understanding what users need, learning to design interfaces, learning to validate ideas, learning to launch, and ending up either as a founder or as the head of a product line. The earnings here are more variable and the upside is materially higher if you build a business that finds its market.

You could become an applied researcher, which means going deeper on the science of how these systems work, learning the mathematics behind them, learning to publish, and ending up at a research institution or a frontier lab. This is the longer path and the one with the highest entry barrier. It is also the path where the Caribbean is most under-represented in the global conversation, and the one where each Caribbean woman who arrives counts for more.

You could become a domain expert who builds, which means staying in your existing field, medicine, law, education, agriculture, finance, and bringing AI capability into it. This is, in my view, the highest leverage path for most mid-career Caribbean women starting today. The combination of deep domain knowledge and competent AI building is rarer than either skill alone, and worth more in the Caribbean market than either skill alone.

Pick a direction. Make a six month plan for the next stage. Find a community. Keep going.

A final word.

I have taught this curriculum, in various forms, to several hundred Caribbean students over the past three years. The pattern I have observed is consistent. The women who finish it tend to be the women who started it doubtful, and the doubt itself was the asset. They asked more questions. They did not assume they understood. They were patient with the parts of programming that the men in the class tried to skip past. They built things that worked because they tested them, where the men built things that they assumed would work because they wrote them.

If you have read this far, you have probably been told, at some point, that this kind of work is not for you. The Caribbean's intellectual tradition has heard that kind of telling before and given the necessary answer. The work is for whoever does the work. The curriculum is in your hands. The first month begins on a Monday of your choosing.

Walk well.

---

Kwame Stevens teaches computer science and AI at a secondary school in Spanish Town and runs the Caribbean Code & Curriculum programme. He coordinates the CAIRA Learning Working Group.

Originally published in The Monthly Intelligence Report, March 2025.