A company that searches for an AI engineer in Mexico is usually in one of two positions. It is a Mexican company that wants someone senior who works in its language and its hours. Or it is a company in the United States that wants an engineer one border away instead of ten time zones away. This page is for both.
What I do
I build artificial intelligence systems that run on real data and stay up: agents that use a company's systems, the evaluation that decides whether a new version ships, MCP servers that give an agent narrow and logged access to data, and the data platforms underneath. I am not an agents-only engineer. Half of the work in any of these systems is plain engineering: data, permissions, tests and operations.
What is specific to Mexico
Three things are, and none of them is a slogan.
The clock. Monterrey is on the same hour all year. For a team on US Central Time that means a shared working day. For a team on either coast it means a difference of one or two hours, which a morning call absorbs.
The language. I work in English and in Spanish. For a company in Mexico this removes translation from the process: the people who know the process describe it in their own words, and the tests are written in the language the users write in.
The distance. I work remotely. Nothing I build depends on where I sit, because it lives in the company's repositories and cloud accounts.
What companies ask for
A Mexican manufacturer usually starts from the plant: work instructions, quality records and supplier mail that nobody can search. A distributor starts from the quote, which is where its sales time goes. A corporate group starts from governance: who may let an agent touch which system, and who signs when a version changes. A team in the United States usually starts from a product: an agent that works in a demo and needs tests before it reaches users.
The method under all four is the same. Describe the process, give the agent narrow access to the data it needs, and write the tests that decide what ships.
What I can show
- Evaluation gates in production. A multi-agent system that serves a higher-education group with 700,000+ students, where no version ships without passing its tests. It is written up as a case study.
- Governed access to tools. An MCP gateway on Cloud Run that puts many tools behind one entry with permissions.
- Open source that others use. An Airtable MCP server, public on GitHub.
- Work for Mexican industry. The platform of a lean manufacturing practice in Monterrey, with an assistant that answers only from its own documentation.
- Research. Pre-registered papers on how agents behave, listed under research.
Each of these says what it supports and what it does not. A case study about one system does not prove I can build yours. It shows how I work and what I check.
How hiring works
Write to me with the role or the problem. For a full-time role, the CV has the history. For contract work, I read the process first and answer in writing with what I would build first, what access I would need and what I would not promise. The services pages describe each kind of work with its limits.
Where I work with companies
In Mexico, with companies in any city: the pages in Spanish cover the largest metropolitan areas, each with its own industries. In the United States, the page on nearshore work lists the cities. In every case the arrangement is the same: I am in Monterrey and the work is remote.