Portfolio
The people I’ve backed.
How I met each founder and what caught my attention. I later interviewed several of them on the podcast.
Enzo Cavalie
1 of 17see them all
bem
Antonio Bustamante, Spanish, raised between Madrid, Barcelona and Quito, and Upal Saha.
Two years ago Antonio was arguing almost against agents. He said that shape carried risks for large companies and would not solve their problem. It made it harder for him to raise and to win customers. Today, what is happening with security, reliability and returns is proving him right.
Turns messy files (PDFs, emails, spreadsheets) into structured data and workflows.
Bizmark
Rodrigo Mosqueira and his cousin Oscar Aguilar, from Peru.
Rodrigo and I studied economics together in Lima. I kept telling him he was ready to start something; he kept telling me he had to get more technical first, and at 28 he moved to Chicago to study computer science. When he finally started, with his cousin Oscar, I wrote the first check.
AI agents for manufacturers and distributors: procurement, quoting and production planning.
bizmark.aivia Reach Capital
Braven
Carlos Chávez, Colombian, based in San Francisco. He had sold two companies in Latin America before this one.
I met Carlos on one of his trips to San Francisco. He walked into the US insurance industry as a complete outsider and still reads the market, knows who to talk to, and connects deeply without making it transactional.
An AI operations platform that turns every submission an MGA or reinsurer receives into a pre-analyzed, traceable decision.
Canonical
Luis Cusihuallpa (Cusco), Rodrigo Chaname (Lima) and Esteban Leiva (Bogotá).
My first job was at P&G, so I saw how big brands research their consumers, and I still struggle to see how that turns into better decisions. I told them so on the first call. The speed convinced me. They brought in founders from Nubank, Rappi and Auth0.
Foundation models of human behavior, to simulate how people will react before you launch.
Chofex
Gonzalo Calderón Gómez, CEO, and Josue Anyosa, CTO, both from Arequipa, Peru.
Gonzalo was building systems for a Mexican lender when he saw how much unused data trucking companies were sitting on. As he sees it, 80% of trucking’s problems are in the dispatch office, not on the road. He spent two months in San Francisco to recruit Josue, who was working at Meta.
An AI employee for Mexican trucking companies: it talks with drivers, customers and the back office over WhatsApp and tracks every trip.
Cometa
Andrés Benavides, from Peru, with Gonzalo Gallardo, Mateo Creamer and Valeria Tapia.
My first articles did not come out in Startupeable, which did not exist yet. I sent them to Ruta Startup, Andrés’s blog, to see if they would take them. Years later, at Reach we invested in Cometa when there wasn’t even a deck. We had to ask him for one.
The operating system for private schools: billing, admissions and communication in one place.
cometaedu.comThe episodevia Reach Capital
EFEX
Dimitri Zaninovich and Santiago Bustamante, who met as students at Stanford.
Three banks sanctioned by the US Treasury left almost a fifth of the market open, and EFEX is taking it. In one year they went from moving 150 million dollars to 1.3 billion, selling through a network of nearly a hundred of their own brokers.
Cross-border payments for businesses: global accounts, local currency collections in 31 countries and transfers to 140.
Harbor
Jesús Santos, from Spain, with Cristian Bartolomé, Daniel Elechiguerra and Luis Botin. All four came out of Stanford.
Jesús didn’t want to build an agents company, but something that lasts 30 to 50 years. He asked which companies last that long: insurers. There, incumbents own the data and the distribution, so he went after new risks with no history, like AI operational risk, which insurers were already excluding.
Insurance for the agentic economy: it measures and mitigates the risks of every AI agent a company deploys, and insures what remains.
LabLabee
Samir Tahraoui, Mahfoud Sidi Ali Mebarek and Chen-Do Tran. Algerian founders, based in Lyon.
A French investor friend introduced us. They are telecom engineers obsessed with their own industry, and before AI they built a simulator of how a network is actually run: design, configuration, maintenance. That simulator turned out to be the perfect place to put agents to work.
The AI copilot for telecom teams: test, troubleshoot and run networks faster.
lablabee.comvia Reach Capital
Maslow
Diego Boryszanski, Matías Botbol and Ricardo Migoya, from Argentina.
They came to me, through the podcast. When I saw that Matías had founded Taringa, the largest Spanish-language social network, where I spent my teenage years before Facebook, I said I was in, whatever they were building. It turned out to be a multi-million business run by thirty people.
Benefits, incentives and recognition for teams, across 25+ countries.
MCPJam
Marcelo Jimenez Rocabado, of Bolivian roots, and Prathmesh Patel. Both came from Asana.
Mateo Puña, a podcast listener, introduced me to Marcelo. He was early to MCPs, back at Asana, with his own view of the infrastructure they would need. That view came from his work there and from building the open source himself, to the point where he is now one of its main contributors.
Open-source testing, debugging and evals for MCP servers.
Meridian
Alfonso Velasco, Mexican, raised in Austin, and Sid Vijay.
I met Alfonso through a cold LinkedIn message; I no longer remember who wrote first. He was 23 or 24 and spoke like someone twice that age. With that maturity he went and sold to supply chain corporates, a traditional industry, with very little experience behind him.
Frontier AI agents for manufacturing, supply chain and staffing operations.
Oqoqo
Renzo Viale, from Peru, and Haritha Nair, who was at Windsurf.
Renzo went to my university in Peru, a couple of years ahead of me. He started one of the first hedge funds in the country, in a market with almost no capital markets. Raising money there and returning results is no small thing. It was a bet on the founder, on whatever he builds.
Custom evals and benchmarks that measure how well AI agents use a product.
Sarelly
Rémi Martini, a Frenchman who spent more than ten years at L'Oréal, Coty and Mustela, and the creator Ana Sarelly.
To test the idea, Rémi launched six products at once with the same nopal leather bag maker: with the Club América football team, with fashion brand licenses and with Ana. Only Ana's worked. In its first year, Sarelly passed a million dollars in sales and was already profitable.
A fashion and beauty brand that wants to be the Chanel of Hispanic Gen Z, built with a creator as co-founder rather than as a sales channel.
Sira
Nathan Belaye, born in Ethiopia, and Antonio Chan.
Nathan could sell sand in the desert. You like him, and you end up wanting him to win. Antonio is the other extreme, the engineer who two years ago was already saying «we are three people and fifteen agents». He was one of the first genuinely agent-native CTOs I met.
An AI HR manager for hourly teams: scheduling, time tracking and payroll.
sira.teamvia Reach Capital
Vetto
José André Nunes, from Brazil, and a team that came through DeepMind and Mistral.
A Brazilian friend introduced us, right after José had dropped out of university. What surprised me is who he convinces. Owners of large Brazilian companies backed him from the first version, when it was still called StartCarreras. He sells without selling, and that is how they later reached DeepMind and Meta.
Environments, evals and expert data that improve the models of frontier AI labs.
vetto.aiThe episodevia Reach Capital
YaVendió
David Tafur, Sebastián González and Terry Cruz, building a Peruvian company from Brazil.
David is one of the most optimistic people I know. With Sebastián, who is a commercial machine, and Terry, who is a serious engineer, they built something that surprised me when I met them: micro businesses paying several thousand dollars a month. That happens because YaVendió is the best WhatsApp seller in the market.
AI agents that sell over WhatsApp and Instagram for ecommerce brands.