Ferran Espuña: Deep Learning Research Engineer, Mathematician and Computer Scientist

About Me

I have a double degree in mathematics and computer science and am currently working in the field of artificial intelligence research. I thoroughly enjoy solving problems of any kind and am very curious, hard-working, and proactive given the right motivations. I work well both independently, as a leader, and as a part of a team.

Contact Info

• Email: ferranespuna@gmail.com
• Location: Barcelona, Spain
• Phone: +34 600 24 69 87

Professional Experience

Barcelona Supercomputing Center | Research Engineer

2023 - Present
As a Deep Learning Research Engineer for Language Technologies, my main work focuses on Large Language Models. So far, my work has involved:

• Building and automating CURATE, a text processing pipeline designed to work in High Performance Computing environments and used to create CATalog, the largest pretraining dataset in Catalan.
• Actively participating in design decisions and writing pretraining scripts for the Salamandra models, a collection of highly multilingual language models pretrained from scratch using the Marenostrum 5 cluster.
• Researching ways to better evaluate the performance of our language models in open generation settings (not multiple choice).
• Opening research lines such as those for alternatives to Transformers (SSMs in particular) and mechanistic interpretability techniques (in particular, Sparse Autoencoders).

Computer Vision Center | Research Intern

2022
Awarded a fully funded research internship at CVC, which also served as an opportunity to develop my Bachelor’s Thesis.

• Research Topic: Application of Topological Data Analysis methods to study the generalization capabilities of neural networks.
• Supervisors: Professor Sergio Escalera (CVC), Professor Carles Casacuberta, and Rubén Ballester.

ChipScope Research Group at UB | Image Processing

2022
Contributed to the European research project ChipScope aimed at creating a high-resolution microscope the size of a computer chip. My work involved:

• Controlling the illumination and camera of the microscope to take image samples from effectively different points of view.
• Aligning and blending the images through computer vision techniques to produce a global view of the samples.
• Implementing wave backpropagation algorithms to remove interference artifacts caused by the small scale of the setup.
• Designing a user interface to make the process user-friendly.

Education

Master’s degree in Advanced Mathematics and Mathematical Engineering | UPC

Expected Graduation: June 2025

• Courses so far: Commutative Algebra, Number Theory, Coding Theory, Cryptography, Combinatorics and Graph Theory.
• Currently working on my master’s thesis Finding Partite Hypergraphs Efficiently. In it, I apply elementary results in extremal graph theory to create a polynomial-time algorithm that finds large k-partite subgraphs of k-graphs, generating a construction that asymptotically matches the best known bounds for existence.

Double Degree in Mathematics and Computer Science | UB

February 2023

• GPA: 9.0/10
• Achieved honors in courses like Linear Algebra, Algebraic Equations, Complex and Real Analysis, Differential Equations, Advanced Algorithms, and my Bachelor’s Thesis, among many others.

Publications

A CURATEd CATalog: Rethinking the Extraction of Pretraining Corpora for Mid-Resourced Languages

Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation, European Language Resource Association and the International Committee on Computational Linguistics.

Finding Partite Hypergraphs Efficiently — Ferran Espuña (2025).

Preprint submitted to Electronic Journal of Combinatorics. arXiv:2508.10641 [math.CO]

Personal Projects

• Developed a neural network from scratch, enhancing understanding of calculus, linear algebra, and gradient descent.
• Created visualizations for various 2D fractals and physical simulations.
• Designed a framework for 3D visualization of polynomial zeros through raytracing, applying my knowledge of Computer Graphics and Differential Geometry.

Skills

• Programming Languages and Technologies:
  • Python (NumPy, Pandas, TensorFlow, PyTorch, NeMo Framework)
  • Slurm
  • C/C++
  • Bash
  • Java
• Languages:
  • Spanish (native)
  • Catalan (native)
  • English (C2 / Proficiency)

Certifications

• Stanford University: Machine Learning Specialization
• IELTS Academic: 8.5/9 overall band score
• Driver’s License: B permit in Spain