Builder · Argentina
I build things
from zero to working.
Backend engineer with a product instinct. I design systems, write the code, and think about what the thing actually needs to do — not just how it runs.
Building in public, one system at a time.
Who I am
I like building systems that survive reality.
I'm a backend engineer and product-minded builder from Argentina. Over the years I've worked on systems involving scraping, distributed workers, pricing intelligence, SEO-scale platforms, NLP pipelines and large amounts of messy real-world data.
I was part of the early team at Mediarails — a startup later acquired by Impact.com — where I learned what it really takes to build systems that create business value, not just technical complexity.
What drives me most is the process of taking an idea from ambiguity to something usable, reliable and alive in production.
What I do
I build systems that ingest messy, inconsistent data from many sources and turn it into something reliable and useful. That means designing the pipeline, making the hard calls about ambiguous inputs, and shipping something that actually works under real conditions.
I'm comfortable going from idea to working product — not just the engineering layer, but understanding the problem, shaping the solution, and iterating until it makes sense.
How I think
I've spent years turning messy, inconsistent data from dozens of sources into something actually useful. That experience shapes how I approach everything — I've learned to trust nothing at face value, make explicit decisions about ambiguity, and build systems that degrade gracefully instead of breaking silently.
Selected Work
Built end to end — a drop-in Authentication as a Service for indie hackers and startups. One script tag replaces hours of OAuth setup: Google, GitHub, Passkeys, RBAC, MFA and webhooks included. Designed as a cheaper, simpler alternative to Auth0, with a working demo, dashboard, and live pricing.
A SaaS tool that helps websites become visible to AI bots like ChatGPT, Claude and Perplexity. Installs via a single JS snippet — the service analyzes the page, generates a structured knowledge base with semantic contracts and JSON-LD tags, and serves it back to LLM crawlers. Think of it as SEO, but for the AI era.
An ActiveRecord-inspired ORM for Python — built from scratch to understand how the abstraction works, not just how to use it. Convention over configuration, full CRUD, and a clean query interface. The kind of project you build when you want to know what's really happening under the hood.
An ongoing technical investigation into whether a competitive brand protection pipeline is viable as a bootstrapped product. Monitoring newly issued domains via Certificate Transparency logs, combining lightweight heuristics with deeper inspection — rendering sites to extract metadata, content signals, and visual identity elements. Building the infrastructure to find out if it can work at the level of established players.
Price intelligence system tracking 20k+ products across multiple sources. Designed prioritized update pipelines to maintain data freshness under constrained resources. Distributed workers and scheduling strategies to balance real-time needs with cost and performance at scale.
SEO-scale platform indexing over 1M companies. Designed efficient content generation and delivery strategies to serve large volumes of pages while minimizing infrastructure costs and maintaining high availability.
NLP pipeline to extract entities, sentiment and relationships from news data, building a structured knowledge graph for downstream analysis. Transformed unstructured text into connected, queryable data enabling higher-level insights at volume.
Background
After the acquisition by Impact.com, continued contributing to MVPs and proofs of concept, evaluating new tools and technologies for the product roadmap.
Built the full stack from scratch: scraping infrastructure, distributed workers, prioritized update pipelines, and the product layer on top.
A good example of making pragmatic tradeoffs based on real usage rather than theoretical correctness.