AI projects for teams
I take on a small number of freelance AI projects from Madrid, for teams in Spain and across Europe. Remote is the default. Most projects are agents, retrieval over internal documents, or internal tools that have to fit a process the team already runs.
What I build
- Agent workflows with a clear trigger, a named owner, and a fallback for when the model fails.
- Retrieval that returns the document, page, or record behind each answer.
- Internal tools that remove one specific manual step.
- Integrations with the systems the team already uses.
- The deployment around all of it: evaluation, logs, permissions, and failure handling.
- Workshops that end with a decision about whether to build.
Where I have done this before
- At AILY LABS, production LLM agents and a Graph RAG service over factory manuals for a Fortune 500 pharmaceutical company.
- At Multiverse Computing, technical pre-sales and proofs of concept for CompactifAI with AWS, NVIDIA, and Dell, including an on-premises GPU deployment for a defense-sector customer.
More in selected work. For regulated settings, see AI for pharmaceutical operations.
How a project usually runs
- Pick one workflow and name the person who owns its outcome.
- Map the inputs, systems, decisions, and what a mistake costs.
- Write examples that separate a useful result from a confident wrong one.
- Build the smallest version that answers the real question.
- Add deployment, logs, a fallback, and a handover so the team can run it without me.
Sometimes step 2 shows that the process should stay a plain rule with no model in it. When that happens I say so, and the project gets smaller.
Contact
Email hello@felipebasurto.com with the workflow, the constraint, and what a useful result would look like.