Víctor Busqué

About the person behind the notebook

Víctor
Busqué.

Software engineer in Barcelona. I build the AI platforms and agents other engineers depend on — and an idea only becomes mine once I've made it run.

● Barcelona 8+ years NLP & Gen AI Agent systems

A little context

I like ideas that move.

I'm Víctor Busqué, a software engineer in Barcelona. For the last eight years I've worked on NLP and generative AI in production — LLM gateways, agent runtimes, and platforms that hundreds of engineers open every morning.

I'm an avid visual learner. I love reading, and I do plenty of it — but an idea stays foreign until I can see it move. A simulation, a figure running the real process, an animation that shows the thing clearly: that's when it clicks.

How I turn
an idea into understanding.

Reading is usually where the investigation starts. It isn't where it ends.

01 / READ

Read it.

FOLLOW THE LOGIC

02 / SEE

See it.

FIND THE STRUCTURE

03 / RUN

Run it.

MAKE THE MODEL REAL

One direction: closer.

Closer to the systems underneath the product. Closer to the machinery that has to keep working when real people depend on it.

01
The beginning

Natural language, in production

The eight years started here — language systems in production, where a model has to work on Monday morning, not just in a demo.

02
The shift

Generative AI arrives

When generative AI landed, I moved to building with it end to end: prompts became products, and evaluation became engineering.

03
Infrastructure

Platforms other engineers build on

The work became infrastructure — an LLM gateway and agent runtime that hundreds of engineers depend on: their defaults, their guardrails, their on-call nights.

04
Today

Agents at scale

Today, at a NASDAQ-listed fintech born in Uruguay: agents automating real work — where scalability, robustness and security aren't optional.

The boring numbers are actually interesting.

Scale is the honest part of the story. Once a system gets large enough, architectural decisions stop being diagrams and start becoming consequences.

Engineers served 0+ Engineers using the AI platform.
Agents 0+ Agents automating real work.
Agent invocations / month 0+ Autonomous and assisted executions.
Requests / month 0M AI infrastructure traffic.
Tokens / month 0B Language-model traffic flowing through the platform.

The whole site in one sentence

If I can't make it run, I can't explain it.

That's probably the simplest description of how I learn. I don't really trust an explanation until I can construct some version of the thing myself and watch what happens. Understanding is something I like to earn.

03 / This site

A place for things worth looking at twice.

Everything published here is that loop, left running in public. One idea per post, told as a visual story — because a page that moves is the form I understand best.

Engineering is the core. Curiosity sets the edges. If a subject deserves a slow walk, it can become a post.

Walk through the notes
Want to talk?

Say
hello.