Live
Daily urban carpooling in Buenos Aires. Flat prepaid fare that does not spike at rush hour, matching passengers with drivers who were already making that trip.
Co-founder · product and commercial
Buenos Aires, Argentina · open to remote AI product roles
AI Product Manager
who still ships the code.
Most of the systems we live inside were never designed badly on purpose; they just got old and nobody touched them. A car driving to work with one person in it. A kid in a classroom learning to memorize things that are one prompt away. Those are the ones I go after.
I do AI product for a living: agent platforms, automation, data. The rest of the time I build products of my own, with Claude Code in the loop for most of the build.
- Live products built
- 5
- Contributions, last year
- 3,168
- Years in software
- 9+
- Years building solo
- 3
Work
Products I founded or co-founded. All of them are live, all of them are mine to keep running.
Live
An AI tutor for kids and teens in LatAm. Nova gets to know a kid, builds missions around what they already love, and never hands over the answer.
Co-founder · product
Live
A discovery platform for early stage projects. Builders publish what they are working on and find co-founders, investors and collaborators before there is anything to demo.
Founder
Live
An AI product management platform. Specs, roadmaps, sprints, research and audits for PMs, founders and small product teams, built on Claude.
Founder
Live
Custom AI agents for small businesses: WhatsApp, web and voice, wired to the data they already have, with a dashboard on top.
Founder
Rise of Hedera
A desktop game with on-chain assets on Hedera. Solidity contracts, a Tauri client and a sprite pipeline, mostly an excuse to go deep on the chain.
Solo project
In construction
A job-application product for LatAm candidates applying locally and abroad: tailored CVs per posting, tracked applications, and eligibility checks up front instead of after you've already applied.
Founder
Also built and shipped: ComoVenimos, a community for tech and startup people, with a blog and a following graph; Valmidor, daily gamified challenges with rankings and a WhatsApp loop; and Fiteando, a social app for finding a training partner. Some are paused, some ran their course. The six above are the ones I still run.
Craft
What I am genuinely good at, and what I am not.
AI agent products
Spec and ship agent systems end to end: trigger detection, data retrieval, multi channel interaction across chat, WhatsApp and voice, and action execution. The hard part is never the model, it is deciding what the agent is allowed to do.
Agentic coding
Claude Code and Cursor as daily drivers. I write production features, put them through review and ship them, which means I do not just prioritize work, I move it.
Automation and internal tooling
n8n, Activepieces, Retool, Zapier, Make. Building the boring machinery that hands a team its hours back, and the process that keeps it from rotting.
0 to 1
Standing up a function, a team or a product where none existed, then designing the way it runs without me in the middle of it. I have done this from zero more than once.
Quality and release engineering
Six years in QA and release automation before product: test design, on-premise deployments for enterprise clients, pipelines in Python and Jenkins. It is why I only trust systems I can test.
Data and BI
SQL, Tableau, Metabase. Dashboards people actually make decisions on, not screenshots for a monthly deck.
Stack
- Python
- SQL
- Claude API
- Claude Code
- Cursor
- n8n
- Activepieces
- Retool
- Zapier
- Make
- Supabase
- Vercel
- Coolify
- Tableau
- Metabase
- Jenkins
- Hedera
Not my thing, at least not yet: production React and TypeScript at scale, Kubernetes and distributed systems, training ML models. I would rather tell you now than find out together in week three.
Skills
14 open Claude Code skills. No hardcoded stack or preferences: each one asks once, saves the answer in your repo, and says so instead of guessing when that file is missing.
project-setup
Run this first. Reads what it can from your repo, asks the rest with the evidence in front of you, writes the shared profile.
dev
Implements a feature or fixes a bug following your repo's own conventions. Reports what it assumed and what it left out.
ship
The full loop for a feature: plan, implement, validate, review, deploy. Chains the other skills together.
feature-agents
Same as ship, delegated to five subagents with narrow tool access: the one writing the spec can't write code. For changes too big for one context to hold.
test
Runs your validation pipeline and reports honestly: what passed, what failed, what got skipped. Never marks green something that never ran.
review
Code review of a diff. Every finding anchored to a real file and line, with the concrete failure scenario spelled out.
deep-review
Interactive architecture review. Every issue comes with options, effort and risk; you decide the priority, not the skill.
ux
UI audit: accessibility, responsive behavior, interaction states, and consistency with your own design system.
product
A business review: does it solve the real need, do the gates hold, is the flow complete, can people actually find it.
deploy
The release pipeline. Separate permissions for commit, push and deploy, and it never ships a build that hasn't passed.
deploy-qa
QA after the deploy, against what's actually live. Picks its depth by blast radius instead of running the same checklist every time.
decision-log
A log of the decisions you can't recover by reading the code: why that limit, why that model, what got ruled out and why.
design-system
Sets up a design system where there is none, and enforces the one you already have: four gates that fail a check instead of surviving review.
community-manager
Ready-to-publish content for X, LinkedIn, Instagram and more, on your own brand profile and funnel. Never invents metrics, testimonials or social proof.
Activity
Most of the code lives in private product repos, so the graph is the honest part. The repos below are what is public.
What is public on GitHub is the short list below. The product code lives in private repos, which is where most of that graph comes from.
Let's talk
Open to remote AI product roles, and always up for a conversation about agents, automation, or why the commute is still broken. LinkedIn is the fastest way to reach me.