Forward Deployed Engineer

Azumo builds and operates production AI systems for companies ranging from seed-stage startups to Meta, delivered through small teams working inside each client's environment. We are hiring Forward Deployed Engineers: the engineer embedded in a client's team who owns the system and whether it works, from the first conversation through to whatever breaks after release. The role is fully remote across Latin America, aligned to your client's working day.

Azumo has shipped production AI & Software since 2016, and on these engagements there is no analyst writing the requirement, no QA phase waiting to catch what you missed, and no architect approving your design. Those calls are yours to make and yours to answer for. If you have only ever worked with those roles in place, this is not the seat to start in. We have other openings that probably fit you better.

Where this role sits

You are the engineer the client works with. On most engagements you are the only one, sitting inside their team, with their repository, their constraints and their deadlines.

Your lane says what you build: software, data pipelines, models, or AI systems in production. This role says how you are accountable for it. We hire Forward Deployed Engineers out of all four lanes.

One question places the boundary: when the client is unhappy, "was it the wrong feature, or a feature that does not work?" As the technical lead on the account, anything that does not work as expected is yours.Not quite your profile? Check our other openings:

• If you own how an AI system behaves in production — AI Engineer

• If you build the pipelines and retrieval layer models depend on — Data Engineer

• If you decide what to measure and which method answers it — Data Scientist

• If you ship product software with agents in your toolchain — AI-Augmented Software Engineer

What you will build

• The whole thing. Discovery, design, build, verification, release, and the incident afterwards. Not a slice of it handed to you already specified.

• The work nobody assigns. The access nobody granted, the dataset nobody can explain, the integration everyone assumed was someone else's. Unclaimed work is yours by default.

• Saying no to the client directly yourself. Not escalating it, not absorbing it quietly. A reason, an alternative, and the standing to be believed.

• Verification. You demonstrate that the acceptance criteria are met. There is no testing phase behind you.

• Reproducible delivery, run with agents. Specification and tests first, agent context committed to the repository, gates on every change. One engineer carrying a whole delivery only works when the process is reproducible.