AI engineer

About the role

You will build the AI systems our clients run every day: assistants grounded in their documents, agents that act on their tools, and private models on their own servers.

What you will do

• Design and build agents and assistants with retrieval over client data, with access that follows each person’s role
• Build evaluation sets with the client, and measure quality before and after every release
• Choose the right model per task: frontier, open-source or on-device, in the cloud or on the client’s own servers
• Deploy on Azure, Google Cloud or private infrastructure, with monitoring and cost tracking
• Integrate with Microsoft 365, CRMs, ERPs, databases and building systems through APIs and tool connectors
• Write the guardrails: what the assistant may do on its own, and what waits for a person’s yes
• Work directly with clients to turn a problem into a working system, and explain trade-offs in plain words
• Contribute to our own products, where the same ideas run for many customers

What we look for

• Two or more years building with LLMs or machine learning in production
• Strong Python; comfortable with APIs, queues and databases
• Experience with retrieval, evaluation and prompt design
• You already use AI tools in your daily work, and know when to check what they produce
• Clear written English; Spanish or Arabic is a plus
• You explain technical choices to people who are not engineers

Nice to have

• Running open-source models on your own hardware
• Speech: transcription or voice assistants
• Experience in a regulated sector: health, finance, public sector
• Security basics: secrets, permissions, audit logs

Who you work with

You work in a small delivery team with a project lead, a full-stack developer and the client’s own people, with senior AI engineers reviewing design and code.

A typical week

• A short planning call with the team and, often, the client
• Building and testing, most of the week, with pull requests reviewed the same day
• An evaluation run: what got better, what got worse, and why
• A demo of working software at the end of the week

What success looks like

• Assistants that people actually use, with answers they trust
• Quality measured, not guessed, before anything reaches users
• Systems that keep running and cost what we said they would

What we offer

• Fully remote, anywhere, with working hours that overlap with Madrid
• Real work from day one: client systems and our own products
• Senior people beside you, with reviews that help you grow
• AI tools as part of everyday work, not a side project
• A paid practical case in the hiring process, and a clear answer within a week