Hey, I'm Jorge!
Software engineer, Growth Engineer and startup founder.
A brief timeline
- 2022–2023First steps: community manager at FTT DAO and data analyst at Medtronic, with Python, Power BI and an OCR for tender documents.
- 2023–nowFounded The Deep Hub, a machine learning publication on Medium: 570+ articles and ~100 authors.
- 2024–2025Data scientist at Axpe Consulting (BBVA, Cosentino, Sabadell) and Founding AI Engineer at xFractal.
- 2024–nowFounder: CTO of Grupo Cardete Ventures and creator of Zapic, AI agents for social media.
- 2026–nowGrowth Engineer at Bit2Me: experiments across acquisition, activation and retention, plus the technical infrastructure behind them.
Also: athlete, traveller and chess player.
Latest
All postsShelves
All topicsHand-picked paths through everything I've written.
My favourite topic. From how an image is formed to detecting objects with YOLO, through interpolation, HOG, Haar cascades and ConvNeXt.
Concepts that took me a while to understand and I wanted to explain well: teacher forcing, state space models, explainable AI and what came before CNNs.
Neural networks from the inside: a complete guide to CNNs, ConvNeXt and the three ways to build models in TensorFlow.
The practical side: why Python is slow, Tkinter interfaces, scraping to build datasets and a Twitter bot with OpenAI.
Series
Some things I write have several parts or follow an order.
Two parts on classic computer vision: histograms of oriented gradients (HOG) and Haar cascades.
An end-to-end project: training the computer vision model and deploying it with Docker, Flask and Azure.
A guided path: how images are created, interpolation, CNNs, ConvNeXt and YOLO.
The data science and machine learning publication I founded on Medium, with 570+ articles from ~100 authors.
Projects
All projectsThings I have built along the way.
Personal portfolio
This site: a portfolio and blog built with Astro and Markdown content, custom pixel-art icons and a playable map to explore it.
X-Ray · HackSpain
Embat challenge: a 0–100 financial health score for SMEs built from their treasury data, with a 3-month forecast, alerts and a scenario simulator.
Colosseum
Top 10 at the Colosseum Solana hackathon with xFractal, a trading platform powered by AI agents.
Home Improvement Scraper
Real-time price comparison API across Leroy Merlin, BricoDepot and Bauhaus, with distributed workers and caching.
Cell Vision
Detection and classification of peripheral blood cells: OpenCV segmentation and a fine-tuned ConvNeXt (98.5% accuracy), served in a Flask app.
Parking Detector
Detects free parking spaces from video with a TensorFlow CNN, deployed with Docker and Azure.