17°

// Senior Solutions Architect

Systems you can actually follow

Solutions architect in the financial sector. My job is drawing how systems that move money connect to each other, and defending those decisions in front of people who ask good questions. What you see here is the other half: things I test on my own swarm of machines, with the numbers and the failures exactly as they came out.

See projects Read the blog Try Agatha

Frame 02 · What I do

The blueprint first, the code after

I draw how a system is going to work before it exists, and then I answer for why it was built that way. What I cannot test at work I test at home: a local GPU, workers and a VPS where the numbers get measured for real.

Open Finance

FAPI 2.0, OAuth 2.0 with OpenID Connect, mTLS and consent management. Chile’s Open Finance System, with demos of all 4 roles the regulation defines, inside the regulatory sandbox.

Banking cloud

Moving a banking core and its digital lending platform to the cloud without switching the business off: what it looks like at the end, how it connects to what is already there, and in what order. GCP and AWS.

Architecture

DDD, hexagonal, CQRS and event-driven. The blueprint for everything else, held up by diagrams.

Backend and systems

API-first microservices in NestJS, Spring Boot and FastAPI; queues, Pub/Sub and systems that hold up under load.

Networks and low level

OSI, TCP/IP, TLS and how a packet actually travels. PhantomRelay came out of that obsession.

AI and generative video

LoRAs trained on my own 16 GB GPU, cloned voice, and an agentic graph that directs video from script to render.

This is not an animation: it is your own connection and the real path this page travels. Zoom into the map to follow the hops. Read on the server, with no JavaScript and no cookies.

Frame 03 · Blog

Latest writing

Go

Go 1.27 brings portable SIMD: it ties with NumPy out of cache and loses inside it

I measured Go 1.27's experimental simd package against NumPy on a real task. They tie when the corpus does not fit in cache, and NumPy wins by 2.4 times when it does. Along the way I nearly published two false comparisons, and those are the useful part.

Read
Models

TabPFN and TabICL against tuned XGBoost: the model that does not train won on fourteen tables out of fourteen

The claim behind TabPFN and TabICL is that they predict on a table without ever training on it and still beat tuned boosting. I measured it on fourteen datasets from the Grinsztajn benchmark, with the same split and the same clock for everyone. The one that does not train wins, the advantage holds up to 32,000 rows instead of breaking, and the most-cited model can no longer be downloaded without an account.

Read
Models

I ran AIFS 2.0, ECMWF's weather model, on my desktop card and landed 0.45 °C from the official forecast

ECMWF published the weights of its AI forecast model. It weighs under a gigabyte and runs on a desktop card: 48 hours of forecast in 108 seconds and 5.77 GB of memory, or 33 minutes if you have no GPU. I installed it step by step, compared it against the same centre's operational physical model, and wrote down the two stumbles that nearly made me publish nonsense.

Read

Frame 04 · Track record

The road here

  1. 2008

    First lines

    Web development for retail and real-estate brands. That is where I learned to ship.

  2. 2017

    Fintech from the inside

    Transactional crowdfunding platform and payment-provider integration.

  3. 2020

    End to end

    IoT with Arduino and AWS IoT Core on Kubernetes, from the sensor to the mobile app.

  4. 2021

    Tech Architect Leader

    Solutions on GCP and DDD-based architectures.

  5. 2024

    Solutions architect

    Architecture in the financial sector: banking core and digital lending to the cloud.

  6. 2025

    Open Finance

    Active work on Chile’s Open Finance System under Fintech Law 21.521: FAPI 2.0, mTLS and the regulatory sandbox.

  7. 2026

    Research of my own

    Agatha with its own paper, adaptive compression in Rust, and an agentic video studio.

Want to work together?

Write to me about architecture, backend or technical content work.

Write GitHub