Matias Martinez is a Software Engineer and Researcher based in Barcelona, Spain.

Personal Sites:

Hello! I am Matias Martinez, a software engineer and AI-Data Scientist. I obtained a PhD in computer science.

Key Expertises and technological interests:

  • Software Development & Testing

    – I focus on building AI-based software, assuring its quality across the lifecycle. Expert on software Testing and DevOps. I developed several open-source apps including Astor.

– I am a teacher in a Master course on MLOps at the Universitat Politecnica de Catalunya (Barcelona, Spain). I teach how to build, test and deploy AI-based systems.

  • AI and Machine Learning:

    – Experience with: Transformers, Experimentation with large language models, Deep learning, Learning to ranking, Statistical models for code analysis, Hyperparameter optimisation, Code translation, and more.

    – Research record on AI-based Software Engineering and Software Testing. Check my Google Scolar profile.

    – Contribution done to open-source large-language models (LLMs), e.g. to CodeLlama by proposing new features (see PR-143) and bug-fixings (see PR-168).

    – I participated in the hackathon HACKBCN AI edition (Barcelona, 2024) and my project won the award “Best developer solution” HackBCN.

Experience

Universitat Politécnica de Catalunya (UPC), Barcelona, Spain

January 2023-Present

Researcher in AI-based Software Engineering

University of Polytechnique Hauts-de-France and CNRS, France

Sept. 2016 - 2022

Associate professor

Inria & Univ. of Lille, France

Oct. 2011 - Oct. 2014

PhD student

Baufest, Argentina.

March 2009 - Sept. 2011

Software Engineer, Java & Web developer, architecture consultant

EDSA, Argentina

Sept. 2006 - Dec. 2007

Java developer

Education

PhD in Computer Science

Université de Lille (France)

Software and System Engineering (BA & Master)

Universidad del Centro de la Prov Buenos Aires -UNICEN-(Argentina)

Competitions

HACKBCN AI edition (Barcelona)

2024

Best developer solution

Ranking of Most Impactful Early-stage Software Engineering Researchers

2013-2020

Second place (worldwide)

Ranking of Most Active Early-stage Software Engineering Researchers in Top-Quality journals

2013-2020

Nineteenth place (worldwide)

Tools & technologies

  • RAG (LlamaIndex, LangChain)

  • ML tools (PyTorch, Pandas, HuggingFace libraries)

  • CI/CD (Github Actions, Jenkins, Travis)

  • Deployment (Docker)

  • Services (Spring Boot -Java-, FastAPI, Django -Python-)

  • LLM inference (vLLM, HuggingFace accelerate)

  • Energy and Resource measurement of AI system (Nvidia-smi, IntelRAPL)

Skills

  • Agile methodology
  • Scrum Master certification
  • Java & Python programming
  • Software Architecture
  • Software testing
  • Continuos integration & deployment
  • Databases (Relational, Non-relational, Vector)
  • Green AI and Software sustainability

Open-source tools developed

Research topics:

  • Automated program repair: – Neural Network based automated program repair – Overfitting analysis
  • AI for software engineering:
  • Software Engineering for AI
  • Green/Sustainable Computing – Energy consumption during training and inference of AI
  • Software evolution:
    – Structural Differencing of code – Evolution patters and trends – Software migration (e.g., Java to Kotlin)
  • Software Security: – Open-source supply chain attacks

Contact: