Two Puerto Rican engineers who spent years building AI tools in Silicon Valley have quietly returned to San Juan to launch a startup that could reshape how companies use artificial intelligence—starting with their own island.
Javier Morales, 34, and Ana López, 29, co-founders of IslaIA, are betting that AI’s next frontier isn’t just in the U.S. or China, but in the global south. Their platform, launched in beta this month, lets small businesses in Puerto Rico train custom AI models using local data—from agricultural yields to hurricane evacuation routes—without needing Silicon Valley-level budgets. “We saw how AI was being built for the 1%,” says Morales. “We’re building it for the 99% who’ve been left behind.”
Why this matters now: Puerto Rico’s tech scene has long struggled with brain drain, sending its brightest minds to Silicon Valley while local industries lag. But IslaIA’s approach—combining open-source tools with territorial incentives—could turn the island into a testbed for “decolonized AI,” a movement gaining traction among developers frustrated with Western tech dominance. “This isn’t just about Puerto Rico,” says Dr. Amara Diop, a Stanford AI ethics researcher. “It’s about proving that AI can serve communities, not just extract from them.”
The brain drain paradox: Puerto Rico has produced more than 60,000 computer science graduates since 2010, yet fewer than 5% stay on the island, according to the Puerto Rico Data Hub. Morales and López are part of a rare reverse migration, joining others like TechPort’s 2025 cohort, which saw a 30% increase in returnees citing “local impact” as their primary motivation. “The talent is here,” says López. “We just needed the right tools to keep it here.”
How IslaIA works—and why it could spread: Unlike cloud-based AI services that require massive datasets, IslaIA’s models run on edge devices (think Raspberry Pi clusters) using federated learning—a technique that keeps data local. For example, a coffee farmer in Yauco can input soil moisture readings from his fields, and the AI suggests optimal harvest times without sending raw data to a server in California. “This is AI sovereignty in action,” says ITU’s AI for Development report, which highlights Puerto Rico as a case study for “territorial data autonomy.”
The Silicon Valley connection: Morales and López cut their teeth at Google and Microsoft, respectively, but left after internal debates over ethical AI stalled. “We kept hearing, ‘Let’s wait for the data to speak,’” says López. “But whose data? And who gets to decide?” Their frustration mirrors a growing schism in tech, where 42% of Latinx engineers in the U.S. say they’re “actively seeking to work on projects with direct community impact,” per a 2025 Gartner survey.
What’s next for IslaIA—and the global south: The startup has already secured $1.2 million in seed funding from VenturePR and the Inter-American Development Bank, with plans to expand to the Dominican Republic and Colombia by 2027. But scaling requires solving a bigger problem: how to make AI tools accessible without relying on U.S. cloud infrastructure. “The real innovation isn’t the code,” says Morales. “It’s the business model—proving that AI can thrive outside the usual suspects.”
A test case for decolonized tech: IslaIA’s model aligns with a broader movement to decentralize AI development, from MOSS’s open-source alternatives to African tech hubs pushing for local data ownership. “Puerto Rico is uniquely positioned,” says PRTech Alliance CEO Carlos Rivera. “We’re not a developing nation, but we’re not the U.S. mainland either. That middle ground is where the real experiments happen.”
The takeaway: IslaIA isn’t just about keeping Puerto Rican talent on the island—it’s about redefining what AI can do when built by and for communities that have been excluded from the conversation. As López puts it: “We’re not building AI for Puerto Rico. We’re building it with Puerto Rico.” For businesses, policymakers, and engineers watching, the question isn’t whether this will work—it’s how fast it can scale.
What would your community need to build AI tools that actually serve it? Share your thoughts in the comments.
