Senior Software Engineer · Buenos Aires, Argentina
I build products end to end, from Go services to React interfaces, for millions of people.
8+ years shipping software at scale: four years at Mercado Libre on subscriptions and shipping, and now the retail platform that PepsiCo, AB InBev and Kroger use in their stores every day.
Lately I've been working on AI agents, and the unglamorous parts that make them reliable: harnesses, evals and guardrails.
Software I've built is used at
Selected work
FORM (GoSpotCheck) is the platform field teams at some of the biggest consumer brands use to run their work in stores: tasks on a mobile app, shelf and cooler photos read by AI image recognition, and reports that tell a brand what's actually on the shelf. I'm a senior engineer on the team building it.
GoSpotCheck by FORM
A rep takes a photo of a shelf; the platform finds out-of-stocks, prices and planogram issues, and turns them into the next task for the team.
Latin America's largest e-commerce and fintech company. I worked on features that every buyer touches, on a platform with 220M+ users.
Subscription program
Mercado Libre's subscription: free shipping, discounts and streaming in one monthly plan. I helped build it from the start. It launched in Mexico in 2023, then Brazil, Chile and Argentina.
Checkout
Reworked how buyers type and validate their address. Fewer typos and missing details meant 6% fewer bad addresses, and fewer packages that never arrive.
Tu día de entregas
Buyers pick one day of the week and their orders arrive together on that day. One predictable delivery for the buyer, fewer trips for the courier.
What I work on
01
Getting an LLM to do something once is easy. Getting it to do the right thing every day, on real data, is mostly engineering. I build the harness around the model: well-scoped tools and MCP servers, retrieval when it needs context it doesn't have, evals that catch regressions before users do, and guardrails for when it's wrong anyway.
02
Go services on GCP and AWS, each with one clear job, talking through queues where it makes sense. I've built services that absorb traffic spikes of 100x normal load, split a tangled legacy service into one with distinct error types and targeted retries (70% fewer errors), and made search-heavy pages up to 5x faster by reworking data structures and OpenSearch queries.
03
React and TypeScript for a lot of people: micro frontends and Server-Driven UI on a platform with 220M+ users, and a Chrome extension that has to live inside other companies' web pages without breaking them. At that scale small details count.
04
I like knowing something broke before anyone files a ticket. Better monitoring, alerting and infrastructure helped cut downtime by about 90% on one platform. Parallelizing a test pipeline took CI from about five minutes to under one, which changes how often people are willing to ship.
Experience
Based in Buenos Aires, working remotely with US teams. Spanish is my first language and I work in English every day. I'm currently studying Computer Science.
The long version is on LinkedIn.
– Present
–
–
–
