services

Work with me

I work with teams that need AI systems to actually reach production — and with people who want to build them. Talks, mentoring and consulting, in Spanish or English.

Talks & workshops

For teams, universities and events

Sessions on applied AI that are built from systems actually running in production, not from a vendor deck. I bring real architectures, the decisions behind them and the things that broke — which is usually the part people remember.

  • Production AI and MLOps: what changes between a notebook and a system you are on call for.
  • LLMs and agentic pipelines: where they earn their keep and where they quietly cost you money.
  • Reinforcement learning for decision-making, from my research on portfolio management.
  • Hands-on workshop format available, with a repository the team keeps afterwards.
Format
Remote or on-site · from 45 min to a full-day workshop · Spanish or English
Rate
On request
Ask about this

Mentoring

For data scientists and AI engineers

One-to-one sessions for people who can already train a model and now need to make it someone else's dependency. I have mentored engineers and set quality standards on client teams, and this is the same thing without the consultancy wrapper.

  • Career direction: moving from analysis into engineering, or from engineering into AI.
  • Code and architecture review on your real project, not a toy exercise.
  • Portfolio and CV review from the side of the table that does the hiring.
  • Preparing for technical interviews in data science and AI engineering.
Format
Remote · 60 min sessions, one-off or recurring · Spanish or English
Rate
On request
Ask about this

Freelance consulting

From architecture to something that runs

I take AI systems from an idea to production and stay for the part where it has to keep working. Nearly four years of client delivery in energy, insurance, SaaS, legal tech and medical imaging — and a portfolio of systems I host and operate myself, which is the honest test of whether the advice holds.

  • Architecture and technical due diligence on an AI system before you commit to building it.
  • LLM and agentic pipelines: RAG, structured extraction, evaluation harnesses, LLM-as-judge.
  • Open-source LLMs deployed on your own infrastructure when data cannot leave the building.
  • Predictive modelling, computer vision and forecasting, taken through to deployment.
  • Rescuing a proof of concept that works on a laptop and nowhere else.
Format
Remote, Madrid-based · by project or by retainer · Spanish or English
Rate
On request
Ask about this

Questions people ask first

What kind of work do you take on?
Three things: talks and workshops, one-to-one mentoring, and freelance consulting on production AI. The consulting work is usually one of two shapes — a system that works in a notebook and needs to become something you can run and be on call for, or a model already in production that nobody can quite explain or afford.
Do you work remotely?
Yes. I'm based in Madrid, Spain, and work remotely with teams anywhere in a compatible timezone. For talks and workshops I also travel on-site.
What languages do you work in?
Spanish and English, both fully. Talks, workshops and written deliverables are available in either.
How do you price your work?
Per engagement rather than from a public rate card, because a 45-minute talk and a multi-week consulting engagement are not the same unit. Tell me the shape of the problem and I'll come back with a scope and a price.
What does a typical engagement look like?
It starts with a short call to work out whether I'm actually the right person — I'd rather say no early than bill you to find out. From there, a written scope with a defined deliverable. Talks run from 45 minutes to a full-day hands-on workshop, and workshops leave the team with a repository they keep.
What technologies do you work with?
Python across the board. LangGraph and multi-agent orchestration, LLM APIs with structured output, FastAPI services, and the operational half — containerised deploys, reverse proxying, SSO, Postgres, and CI that ships on push. I also come from a quantitative background: reinforcement learning, valuation models and time series.
Can I see something you've actually built?
Yes, and without creating an account. Warren and Gollum are both live and both have a read-only guest mode, so you can follow the full reasoning end to end. The write-ups on the blog cover the architecture decisions behind them.

Something else in mind?

If your case does not fit neatly into one of these, write to me anyway and we will figure out the right shape.

Send me an email