‹ Projects

FrameX Labs

Partner Product Engineer at an AI-native product studio
Barcelona · 2025 – today · framexlabs.com
Visit site Ask me about it
Role
Partner
Product engineer, not a contractor
Since
2025
Still there
Clients
4
Natura, Prosegur, AJA, Codeworks
Dev cost
−35%
Lovable, Supabase, n8n
Support
−75%
First response time
Launches
2
AI products in 3 months

Preview

FrameX Labs screenshot
The surface
#0B0F17
Near-black with a star field. A studio selling AI infrastructure has to look like it builds the thing, not like it resells it.
The arrangement
Partner, not supplier
I own the problem, not the ticket: discovery, scope, build and launch. That is why Minerva exists.

About

FrameX Labs is a product studio that helps companies become AI-native: automation, data infrastructure and AI agents wired into how a business already runs. I joined in 2025 as Partner Product Engineer.

The title is the point. I am not handed a spec to implement — I get handed a business problem, run the discovery, decide what gets built and then build it. Minerva came out of this: designed, developed, tested and launched end to end, and now in production with commercial partners in Brazil.

Alongside that I run product assessments and consulting for Natura, Prosegur, AJA and Codeworks. Development cost is down 35% using Lovable, Supabase and n8n; support first response time is down 75% through automated WhatsApp flows; and two AI products went from nothing to launched in three months.

Version history

Minerva, end to end
Designed, built, tested and launched inside FrameX.
2025 – 2026
Two AI product launches
MVP and go-to-market in three months.
2025
Client assessments
Product reviews for Natura, Prosegur, AJA and Codeworks.
2025 – today

Notes from the maker

3 notes
Partner, not supplier
Martín
A supplier is handed a spec. I get handed a business problem and decide what gets built, which is the only arrangement where a number like minus thirty-five per cent means anything.
Where the 35% actually came from
Martín
Not from writing code faster. From killing things before they were built. A prototype in an afternoon means you discover it is a bad idea that same day, instead of in sprint review three weeks later.
What AI still gets confidently wrong
Martín
Data models and payment paths. It hands you something that runs and quietly loses an edge case, with total confidence. I review those two by hand on everything, every time. The rest it does better than I do.

You might also like

MinervaPatient portfolio for surgeonsVIEWCeramicosoCurated ceramics marketplaceVIEWBebi Puig-PeyPortfolio for a filmmakerVIEW
Martín Gabisson · Barcelona · mgabisson@gmail.com