Software engineer

I build the
hard stuff.

Machine-learning models, full products, native apps, whole cloud systems. I like the problems nobody wants to touch, and I build them to run fast, ship fast and stay cheap.

Maths and Computer Science · University of Barcelona3 years shipping a production SaaS
How I work

Three things I watch in everything I build.

01

It runs fast

Performance is a feature, not something you bolt on at the end. I care about latency and the shape of the data path.

02

It ships fast

I move quickly without leaving a mess behind. Clean architecture is what lets you keep moving quickly.

03

It stays cheap

Serverless where it fits, caching where it counts, and cloud bills that actually make sense at the end of the month.

Work

Products I've built and run.

Founder & lead engineer · Production SaaS

Comiendo al Mundo

A restaurant-discovery app I designed and built almost entirely on my own: the app people use, the admin panel, public pages, the backend and all the cloud infrastructure. Web and native iOS and Android from a single codebase. A real product, in production, growing.

Next.jsNode.jsTypeScriptGoogle CloudAWSTerraformStripe
Visit site
Comiendo al Mundo admin panel
Comiendo al Mundo app
Lead engineer · Live product

golf28

A live golf-offers platform where I built the whole tech side: subscription billing with Stripe, automated WhatsApp alerts, and a daily best-offer engine that emails leads every morning. We're a team of three.

Node.jsTypeScriptGoogle CloudStripeWhatsApp
Visit site
golf28 landing page
My first venture · Built solo

waitr

Where it all started, and I built the whole thing on my own: an all-in-one order manager for restaurants. Customers ordered and paid from their phone, whether eating in, taking away or getting delivery. It's the foundation Comiendo al Mundo grew from.

AngularDjangoTypeScriptMongoDBDocker
Waitr app
Projects

Things I build for the love of it.

Bachelor's thesis · Machine learning

F1 race simulator

Two custom Transformer models that call Formula 1 strategy live: one picks the pit-stop window, the other the tyre compound, as the race unfolds. Fed by real-time telemetry, tested during real Grands Prix. (That's the live sim, right here.)

95%
pit-stop accuracy
91%
tyre compound accuracy
2
custom Transformers
PythonTensorFlowTransformersPostgreSQLFastF1 · OpenF1
SilverstoneLive sim
A Transformer to predict stock prices

pink-donut

I took the Transformer idea from my F1 thesis and pointed it at the markets: a Python microservices pipeline that predicts where a couple of stocks are heading.

PythonFastAPITensorFlow
A trading bot based on backtested strategies

orange-pumpkin

Backtests showed that buying the 3rd, 4th and 5th biggest companies in the S&P 500 beat the index over the long run. So I built a serverless bot that does exactly that, every month, on its own. It's running.

PythonGoogle CloudAlpaca
Agentic AI for market research

r2k

A set of specialised AI agents that research the market in parallel and reach a conclusion together. It runs on its own, on a schedule, over financial data.

ClaudePythonMulti-agent
stock price
$112.40
Stack

Tools I'm comfortable with.

And if something new shows up in the mix, that's never a problem.

Languages

TypeScriptJavaScriptPythonJavaSQLC++

Frontend & mobile

Next.jsReactAngularTailwindiOSAndroid

Backend & data

Node.jsExpressDjangoPostgreSQLMongoDBFirebaseRedis

Cloud, infra & AI

Google CloudAWSTerraformDockerKubernetesTensorFlowLLMs & agents
Background

Double degree in Maths and Computer Science.

University of Barcelona, one of the toughest programs in the country. My strongest marks were in the parts that build good software.

13.4/14 university entrance examHonours · high school
Software Engineering10
Databases9.8
Software Design9.2
Data Structures9.0
Distributed Systems8.8
Networks8.6

Let's build something real.

Send me a message
LinkedInInstagramEmail