Juan Peñarrieta
AI Leader · Advisor · Board Member
I navigate the decision every company now faces: whether to rent, build, or buy AI.
I evaluated vendors, built AI at Amazon, Meta, and a startup hit by ChatGPT, and sourced acquisitions for an angel fund in LATAM. Rent, build, buy: the same decision, seen from all three sides.
I work with AI every day, and I want to give something back: the stories and lessons I keep finding along the way. My goal is simple: keep the AI knowledge gap from growing, especially in Latin America. I see it as an opportunity: regions that came to the technology late can, this time, get ahead. Learning to choose well what to build, what to rent, and what to buy is part of that.
Learning to rent.
My first job was as a management consultant at Real Time Management, helping companies across Latin America optimize and negotiate their purchase of services, mostly in mining, agriculture, and construction. With GyM, for instance, we evaluated their current suppliers and looked for where to renegotiate a contract or find a better one. I ran auctions where suppliers competed, and I evaluated them in detail: what the company actually needed, and what each supplier was offering. The goal was simple: get the most value out of what the company paid, without paying for what it didn't use. That's where I learned to evaluate suppliers, and years later it served me from the other side of the table.
Learning to build.
At Merlyn Mind I led product. We built a service where schools subscribed to AI skills for their teachers, the first conversational AI for the classroom. The startup raised more than 150 million dollars and the team grew from 20 people to more than 100. We built something that worked, that teachers actually used, and I lived what it costs to keep something like that standing. Then ChatGPT arrived. Within weeks, it changed the way everyone interacted with AI. That scar is behind everything I write today: building is rarely the hard part. Knowing when to build is.
Before Merlyn, at Amazon, I learned the other side of the craft: building well, at scale. I worked on the first multimodal Alexa model and the company's first conversational shopping experience. Today, at Meta, I work on building AI solutions for advertisers. I see the same decision (what to build, what to rent, what to buy) inside one of the biggest companies in the world.
Learning to buy.
With a small angel fund in Latin America, my role was finding the opportunities and putting them on the table. There I learned how a company gets acquired, when it makes sense, and the problems that surface after the deal closes. The capability you buy doesn't always stay; sometimes it walks back out the door. A year later I left for Wharton and stepped away from the fund.
Alongside that, I sit on the board of Global Mapping, a digital mapping company that captures data about the earth and uses AI to help make decisions faster. There, I help build AI on a geospatial archive calibrated over 25 years, information no one else has.
At the end of the day
My main role is being a husband, a father to two active boys, and someone who gives something back to his Latin American roots. The work at Meta, the ventures, the boards (everything else) is built on that foundation.