Fabio Capela

Hi there! I'm Fabio, and I work at LDC in Geneva as a quant trader. I develop trading strategies using quantitative methods. Before this, I worked at Firmenich SA on some AI projects related to flavors, and I was part of starting a company called SamurAI that worked with sentiment analysis.

I enjoy triathlons in my free time - swimming, biking, and running help me clear my head. Family time is really important to me too.

I studied physics and got my PhD from ULB in Brussels, which helps me in my data work today. I also like reading philosophy when I can - it gives me different perspectives on problems.

Schedule a Chat

All Stories

The Soft Logic of Language Models: Mathematical Reasoning in the Age of AI

Large Language Models present us with a fascinating paradox. They can write poetry that moves us to tears, engage in sophisticated philosophical discussions, and even generate code that solves complex...

StackNets StackNets
29 Jun 2025

Graph Neural Networks: Deep Learning on Non-Euclidean Data

Graph Neural Networks (GNNs) represent a fundamental paradigm shift in deep learning, extending the remarkable success of neural networks from Euclidean domains like images and sequences to the irregular, non-Euclidean...

StackNets StackNets
21 Jun 2025

Gauge Fields: The Hidden Geometry Behind Nature's Forces

The concept of a connection stands as one of the most beautiful unifications in mathematical physics, bridging the abstract world of differential geometry with the concrete phenomena of gauge fields...

StackNets StackNets
10 Jun 2025

The Feynman Path Integral: Revolutionizing Our Understanding of Quantum Mechanics

Richard Feynman’s path integral formulation stands as one of the most elegant and profound reformulations of quantum mechanics ever conceived. Unlike the traditional Schrödinger equation approach, path integrals offer an...

StackNets StackNets
25 May 2025

Quantum Gravity and Spin Networks: Weaving the Fabric of Spacetime

Imagine zooming into the fabric of spacetime itself, magnifying it by a factor of $10^{35}$—from human scales down to the Planck length of approximately $1.6 \times 10^{-35}$ meters. What would...

StackNets StackNets
21 May 2025

Why Momentum Works: The Physics of Optimization

Gradient descent is the workhorse of modern machine learning, but vanilla gradient descent often struggles with challenges like saddle points, ravines, and local minima. Momentum-based methods address these limitations through...

StackNets StackNets
16 May 2025