Education

  • Specialization on Bioinformatics and Computational Genomics
    University of Salamanca
    Salamanca, Spain
  • MSc in Visual Analytics and Big Data
    Thesis Title: 'Traffic forecasting in the city of Madrid using Graph Neural Networks'. Advisor: Dr. Ricardo S. Alonso RincĂłn
    International University of La Rioja (UNIR)
  • PhD in Theoretical Physics
    Thesis Title: 'Dissociative ionisation of the hydrogen molecular ion in intense laser fields'. Advisor: Dr. Daniel Dundas
    Queen's University Belfast
    Belfast, UK
  • MSc in Physics and Technology of Lasers
    Master thesis title: 'Interaction of attosecond pulses with aligned molecules'. Advisors: Camilo Ruiz and Ricardo Torres
    University of Salamanca
    Salamanca, Spain
  • BSc in Physics
    University of Salamanca
    Salamanca, Spain

Experiences

  • AI Lead
    Leads a team of ML and MLOps engineers dedicated to building a platform to evaluate a variety of ML models, especially LLMs. The platform can evaluate models both run local or on the cloud. The team is also investigating and developing new metrics to assess LLM's performance in different scenarios. Continues to co-supervise BSc and MSc thesis in collaboration with several Spanish universities.
    HP SCDS
    AI lab
  • Senior Machine Learning Engineer
    Developed a novel tool for digitizing raster images of architectural floor plans (scanned images) into editable vector graphics files (CAD files). This task, an image-to-sequence translation, is performed using CNN for segmenting different objects and transformers to parameterize geometrical objects into a sequence of tokens. Developed a PoC (Proof of Concept) consisting of a digital twin of a biological process for a pharmaceutical company. Co-supervises BSc and MSc thesis in collaboration with several Spanish universities.
    HP SCDS
  • Data scientist / Machine Learning Researcher
    Focused on the implementation of state-of-the-art Deep Learning solutions, mostly based on Convolutional Neural Networks (CNN), to existing ML pipelines. Contributed to the writing of scientific articles and the communication, presentation and discussion of the latest published research in AI to the rest of the R&D team.
    Connect AI team, Nielsen
  • Computational scientist
    Developed computational codes and integrated new scientific modules in existing codes, including their parallelization. Duties also included the management and request of computational time in supercomputing facilities.
    University of Salamanca
    Salamanca, Spain
  • Technical staff
    Job within the eCSE program funded by ARCHER (the UK National Supercomputing Service). Co-designer and developer of the library POPSICLE (Photoelectron SpeCtrum library for Laser-matter intEractions). The software calculates photoelectron spectra for various grid-based solutions of the time-dependent SchrĂśedinger equation (and time-dependent Kohn Sham equations of TDDFT). The library is available on GitHub.
    Queen's University Belfast
    Belfast, UK
  • Early Stage Researcher
    Job within the EU-funded Marie Curie Initial Training Network CORINF. Explored the link between electron correlation and multi-electron dynamics in atoms and molecules exposed to ultra-short laser fields. Worked in the maintenance, development and optimisation of various HPC computer codes, including parallelisation.
    Queen's University Belfast
    Belfast, UK
  • Research staff
    Job within the Consolider SAUUL Project (Funded by the former Spanish Ministry of Science and Innovation). Developed new computational software for the study of attosecond laser pulses and aligned molecules.
    University of Salamanca
    Salamanca, Spain
  • Research Intern
    ECYL grant for training in research and technological innovation for young graduates.
    University of Salamanca
    Salamanca, Spain

Last publications

Patents

  • Initialization of classification layers in neural networks

    Authors: AlmazĂĄn, E., Tovar, J., & de la Calle, A.

    Patent Number: U.S. Patent No. US 11676034 B2

    U.S. Patent and Trademark Office (2023)

  • Methods, systems, apparatus and articles of manufacture to apply a regularization loss in machine learning models

    Authors: AlmazĂĄn, E., Tovar, J., & de la Calle, A.

    Patent Number: U.S. Patent No. US 20220092424 A1

    U.S. Patent and Trademark Office (2022)

Skills

Keywords

Machine Learning (ML), Deep Learning, GenAI, LLM, CNN, GNN, RNN, RAG, MLOps.

LLM Tools

Hugging Face's transformers and datasets, LangChain, llama.cpp, VLLM, SGLang, NVIDIA Tensor-RT, OpenAI API.

ML Frameworks

PyTorch (Lightning), Scikit-learn,TensorFlow/Keras, Flax(Jax).

Data Science

Numpy, Pandas, Tidyverse, Streamlit/Gradio, Biopython, Bioconductor, NetworkX

Visualization Tools

Matplotlib, Seaborn, Bokeh, Altair, Plotly.

DevOps/MLOps Tools

Git, Docker, FastAPI, AWS (EC2, S3), Linux/Unix Administration, HPC schedulers (Slurm).

Languages

Spanish (Native), English (Full Professional), French (Intermediate).