Ailon-Glandt AI engineering · Madrid
AI agents for work that has to be right.
We build production AI systems for companies whose output gets audited — clinical care, tax compliance, regulated operations. Not pilots and demos. Systems that run every day and can show their work.
What we do
Three ways we work with you.
Agentic systems, built to run in production
We design and build agents that complete real work end to end: read the documents, operate the internal systems, follow the procedure, and escalate to a human at exactly the right moment. The hard part is never the model — it's the procedure, the edge cases, and knowing what the agent must never do on its own. That's the part we specialise in.
AI assessment and roadmap
A short, fixed-price engagement that answers three questions about your operation: where AI actually pays off, what it will cost to get there, and what you should ignore. You get a written roadmap you can act on or hand to another team. No obligation to build with us.
Embedded AI engineering
Senior machine-learning depth inside your existing product team for a fixed period — architecture, model selection, evaluation, and the unglamorous work of making a system reliable enough to put in front of users.
Selected work
Two systems, one problem.
Both run in domains where a wrong answer has consequences, and both come down to the same question: what can the agent decide, and what has to reach a person first.
An agentic system that helps doctors treat diabetes patients
Built for owna.health, a technology-enabled chronic care platform that reverses type 2 diabetes through medical supervision, coaching and therapeutic nutrition.
The system works on both sides of the consultation. Patient-facing, it supports people through their programme between appointments. Clinician-facing, it gives the medical team the synthesis they need to make a treatment decision — without reading back through months of raw tracking data.
4× improvement in caregiver efficiency
Clinical agents live or die on knowing their limits. The design problem was never how capable to make the system; it was drawing a hard line between what it can handle on its own and what has to reach a doctor first.
An autonomous gestoría for Spanish businesses
Built as internal R&D, now running real filings in production.
Drop a supplier invoice into a folder and an agent reads it, books it, and files it. Quarterly IVA and IRPF returns (Modelo 303, 130, 202) are prepared and submitted through a guided conversation. Every invoice is registered with the AEAT in real time under Verifactu (RD 1007/2023), with hash chains and QR codes.
What makes it work isn't the model — it's seventeen procedure modules encoding Spanish tax law, a safety ruleset classifying every action as automatic, needs-approval, or forbidden, and strict isolation between legally distinct taxpayers sharing one system.
Spanish companies face Verifactu from 1 January 2027, and the self-employed from 1 July 2027.
How we work
Fixed scope. Agreed price. Something running at the end.
Fixed-scope pilot
Four to six weeks. One workflow, one measurable outcome, a price agreed up front. You end up with something running, not a slide deck.
Advisory, by the day
Architecture reviews, technical due diligence, team mentoring, second opinions on an AI roadmap.
Build and hand over
We write the system, the documentation and the procedure so your team owns it. We don't build dependencies on ourselves.
We're based in Madrid and run client workshops from our own co-working space in the Prosperidad district. Remote works fine too.
About
A small Madrid practice.
Ailon-Glandt Consulting S.L. is an AI engineering practice based in Madrid. Behind it is over twenty years of machine-learning research and engineering — previously at Google, Yahoo! and Pinecone — more than forty peer-reviewed publications, and current university-level teaching in Madrid.
We work with a small number of clients at a time, and the person who scopes your project is the person who builds it.
Contact
Tell us what the workflow is.
And what "working" would look like. If we're not the right fit, we'll say so in the first call.
Madrid, Spain