Washington, D.C., is the capital of the United States. It is known for the federal government, for defense and government contractors, for law and policy, for international organizations and for cybersecurity. In a city like that the first question is not what an AI engineer can build. It is whether he may work on it at all, so this page starts there.
What an engineer outside the United States cannot take
I live in Mexico. Part of the work in Washington is restricted, by law or by contract, to people inside the United States, to its citizens, or to people who hold a clearance. That applies to agencies and often to the contractors that serve them. For such work an engineer in Monterrey may simply not be eligible, and no arrangement changes it. Your contracting officer or your compliance team knows which rule applies. Ask them first.
What remains is wide: commercial products, internal tools that hold no restricted data, systems built on public information, and the work of firms and organizations that share the city with the government without being bound by those rules.
Agencies and contractors, where the work is open
These organizations ask the same things of any system: who acted, under what authority, on which data, and what proves it. An agent has to answer them too. The service is Enterprise AI. No case study on this site comes from government. What transfers is published work for large organizations with processes of that kind. The first piece is the set of evaluation gates that decide what ships in a multi-agent system serving a higher-education group with 700,000+ students. The second is a gateway that places many tools behind a single entry, with authentication in front of it and evidence left after each call. The third is the guide on agent auditability.
Law, policy and international organizations
Law firms, policy groups and international organizations work on long documents and are judged on citing them correctly. What they tend to need from AI is an answer that quotes its source and says so when the source is silent. That is the work under RAG and data platforms. No case study on this site comes from law or policy. What transfers is an assistant I built for a lean manufacturing practice. It answers only from an open corpus of that practice's own documentation: a different field under the same rule, nothing from outside the corpus. International organizations add a second language, and I work in English and in Spanish.
Cybersecurity companies
A security company looks at an agent the way it looks at any new account: what can it reach, and who would notice. The engineering answer is narrow access, with every tool an agent may call defined in advance and every call logged. I build that under MCP servers and integrations. This page does not offer security assessments. It offers agents built so that such an assessment has something to read.
The hours
Washington shares its hour with New York. That puts it one hour ahead of Monterrey from early November to mid March and two hours ahead from mid March to early November, because Monterrey does not move its clock. The work sits in your repositories and accounts, and it is reviewed in writing.
If the work is open to me
Send the role or the problem, and say whether any restriction applies. I read the process as it stands and answer with a first thing to build, the access required and what I would not promise. Senior roles and contract engagements are both possible. I am not in Washington, and where a rule requires someone who is, I am the wrong person.