AI Lead Engineer · Buenos Aires, Argentina
Six years across five sectors — Chile’s first neobank, a 100M-user streaming platform, enterprise travel tech. Compliance systems, event-driven data platforms, sub-200ms money movement. I bring that same discipline to AI: the pipelines, the cost ceilings, and the pager.
AI in production
Anyone can get a demo working. Keeping an agent useful once it touches real systems is a distributed-systems problem — permissions, cost ceilings, latency budgets, what it does when the model is wrong, and someone carrying the pager afterwards. I build agents that do real work inside engineering organisations, on the same production foundations I have operated for six years.
Local agents built on Claude Code, Copilot and Codex, aimed at the parts of shipping that actually cost time: code review, code generation, and deployment. Treated like any other production dependency — standards around them, review gates in front of them, and a clear view of where they stop being trustworthy.
A conversational agent over Buenos Aires rental listings. Paste a listing URL and it answers the only two questions that matter to a buyer: is this priced as an opportunity, and what annual yield would it actually return?
Built on Claude, GPT and self-hosted open-weight models, chosen per task rather than by habit — and tuned for cost and latency, because an agent nobody can afford to run is not in production.
Method
If you can’t measure the output, you’re shipping vibes. The harness comes first, and it stays in CI.
Permissions, pipelines, cost ceilings, latency budgets, fallbacks. That is the system. The model is one component inside it.
Prometheus and Grafana on AI systems, same as any other service. Shipping ends when it’s observable, not when it merges.
Experience
Travel, banking, messaging, streaming, lending. Every one wanted something different: an enterprise where the problem was fifteen legacy provider integrations; Chile’s first neobank, where the user count moved twentyfold underneath me; a platform pushing films and series to a hundred million people. Dropping into an unfamiliar domain and being useful quickly is the part that transfers — the stack has always been the easy half.
Apr 2025 — Present Remote
RyzLabs (Tala)
Lending · mid-size fintech, regulated money movement
Apr 2023 — Apr 2025 Remote
Qubika (Disney+)
Streaming · films and series to 100M+ users worldwide
Apr 2023 — Dec 2023 United States
MediaLab (Kik)
Messaging & streaming · US consumer app
Jan 2022 — Apr 2023 Argentina
Tenpo (Fintech)
Banking · Chile’s first neobank, 100K → ~2M users
Jan 2021 — Jan 2022 Argentina
Despegar (Travel Tech)
Travel & tourism · LATAM enterprise
Speaking
ORT University × Qubika · guest lecture
A guest lecture for engineering students on decentralised governance — what a DAO actually is underneath the terminology, and the practical path from smart contract to something people can use.
Available to speak on:
Capabilities
Beyond coding
Every winter, as high as the lift goes
Medio Maratón, Mar del Plata
Mise en place is just good system design
Also: surfing whenever there is cold water and no signal, and a lifelong, medically inadvisable commitment to Boca Juniors.
Contact
Open to AI Lead and Staff Engineer roles. Remote from Buenos Aires, or relocating for the right team.
Available for new work