Revelo — a platform connecting US companies with full-stack developers across Latin America — has seen demand surge for engineers who can assist with LLM training, co-founder and CEO Lucas Mendes told TechCrunch. The platform hosts more than 400,000 developers and provides streamlined hiring and payment processes for its US clients.
Mendes said the demand surge is driven by the next phase of the AI revolution: post-training LLMs. “There is now a race for data, especially expert-level human data that can actually help LLM students get better at very specific high-value tasks,” Mendes said. “Coding is one of those tasks. Last year we saw a surge from [companies] building foundation models looking for engineers who can work efficiently and provide human data to help their LLM students improve at coding.”
By 2024, LLM-training hires accounted for 22 percent of Revelo’s revenue.
Who’s Buying, and Why Latin America
Mendes said demand typically manifests as companies approaching Revelo to find specialists in specific coding languages to fill gaps in their ongoing post-training pipelines. Revelo is supplying talent to US enterprises including Intuit, Oracle, and Dell — plus “almost every major hyperscale AI provider.”
Revelo is not alone in connecting US companies with Latin American programmers; Terminal, Tecla, and Near are among several pursuing the same goal. But demand for post-training developers is just the latest hiring trend Revelo has ridden since launching in late 2014. Mendes and co-founder Lachlan de Crespigny founded the company because the talent war was so intense they believed a vetted talent network in Brazil would let enterprises find the engineers they needed.
The demand was there. Revelo raised over $48 million from Social Capital, FJ Labs, and Valor Capital Group, and expanded from Brazil across the broader Latin American region. COVID “significantly” widened Revelo’s potential reach: “Suddenly we started getting calls from US companies who realized they could actually have really high-quality distributed teams, and some of those engineers were in Latin America,” Mendes said. “The typical pattern is they hire one or two people who really value the quality-to-cost balance, then say ‘hey, I want more of these people, where can I find them?’”
As companies return to offices, the distributed-work trend has cooled — but Revelo’s momentum has not. Mendes jokes he doesn’t want to go against the tide, but demand for Latin American talent hasn’t declined despite RTO. The reason: nearshoring. Revelo’s developers share time zones with US clients, making collaboration easier than offshore alternatives in Asia or Eastern Europe.
The Acquisition Strategy
Revelo has seen enough demand to acquire five competitors focused on Latin American talent in the past 30 months, including Alto and Paretisa announced in March. “We’re building a global talent backbone for the AI era, and there will be more acquisitions,” Mendes said.
The engineer-as-training-data dynamic represents a genuinely new labor category with distinct characteristics from traditional software contracting. A traditional contractor builds a product; a post-training data engineer produces examples of expert-level reasoning that teach a model to replicate that reasoning. The skill set overlaps — you need to be a competent coder to write good training examples for a code model — but the output is not software. It is judgment encoded as demonstrations, and its value derives not from what it builds but from what it teaches. That distinction matters because it changes the economics: training data has marginal value that decays as the model improves, creating a treadmill where labs must continuously source higher-quality, more specialized expert data to keep improving. The 22 percent revenue share is likely a floor, not a ceiling, because the next frontier — agentic AI that can perform complex multi-step coding tasks — requires even more sophisticated human demonstrations of tool use, debugging, and system design.
The LatAm angle is also a geographic arbitrage story worth understanding. Silicon Valley engineering salaries run $200,000-$500,000+ for senior talent; comparable LatAm engineers with strong English and timezone alignment cost a fraction of that. For AI labs burning billions on compute, shaving 40-60 percent off the human-data line item is meaningful. But the arbitrage has a shelf life: as demand from every major AI lab hits the same talent pool simultaneously, LatAm engineering rates will rise — and Revelo’s five acquisitions in 30 months suggest the company expects the consolidation window to close. The endgame is likely a small number of global talent-platform oligopolies, each with proprietary networks of vetted experts across time zones and domains, competing to supply the human data that the next generation of AI models trains on. Revelo’s 400,000-developer network is the largest in its region, and the five acquisitions signal that it intends to keep that lead. For investors, the interesting question is not whether the AI-training-data market grows — it demonstrably is — but whether platform intermediaries like Revelo capture value or whether AI labs verticalize by building internal expert networks, cutting out the middle layer entirely.
The Bigger Picture: Human Data as the New Oil
The LLM-training boom reveals a structural shift in how AI companies source their most critical input. Pre-training data came from the internet at near-zero cost; post-training data requires human experts — engineers who can write code well enough to teach a model, doctors who can evaluate medical reasoning, lawyers who can grade legal analysis. The supply of such experts is finite, their hourly rates are rising, and the competition for them spans every major AI lab simultaneously. Platforms like Revelo function as the labor-market intermediary for this new economy, and their growth (22 percent of revenue from AI training in one year) is a leading indicator of where AI companies are spending their budgets. The LatAm advantage — timezone alignment with the US, strong engineering education, favorable cost structures — positions the region as the primary sourcing ground for exactly the talent the AI industry’s next phase needs most. The acquisition spree signals that Revelo intends to own that position before the window closes.
