PeopleofSuperintelligence

Researcher

Andrés Carvallo

Also known as Andres Carvallo de Ferari

Andrés Carvallo is a computer scientist whose AI work connects machine learning for biomedical literature, multilingual language technology, and fairness in online discussion. He earned an engineering doctorate in computer science at Pontificia Universidad Católica de Chile and developed methods for explainable screening of medical papers. His research spans Spanish–Mapudungun translation, identity-aware hate-speech analysis, studies of model explanations, and simulations of online conversations. As of October 2026, he is a CENIA researcher and coauthors research on which large-language-model layers matter for building efficient models.

Now at
CENIA

Timeline

  1. 2020

    XLNet for evidence-based medicine

    In 2020, he coauthors a study that uses XLNet to classify biomedical articles for Epistemonikos. The model improves the existing approach by 93% on average F1 and saves physician time in COVID-19 literature screening.

  2. 2022

    Explainable medical paper screening

    In September 2022, he defends “Interactive and Explainable Machine Learning to Improve Efficiency in Medical Document Screening” under adviser Denis Parra Santander. His engineering doctorate in computer science is from Pontificia Universidad Católica de Chile, and his thesis addresses automated, explainable review of medical papers.

  3. 2023

    Spanish to Mapudungun translation

    In 2023, he coauthors an active-learning study of neural machine translation between Spanish and Mapudungun. Its strategies reduce translators’ manual workload by more than 40% in both directions.

  4. 2024

    Postdoctoral fairness project

    In 2024, he is a postdoctoral researcher at CENIA and receives a FONDECYT grant for work in Pontificia Universidad Católica’s computer science department. His project develops multilingual approaches to reduce bias in machine learning and NLP, including language-model bias against identity groups.

  5. 2024

    Identity-group recognition in social media

    In 2024, he coauthors a named-entity recognition system for identity groups in social media. The system identifies attacks and tags the words that name the mentioned groups.

  6. 2025

    Visualizing model attention

    In 2025, he coauthors a study of attention visualization for XLNet’s biomedical-paper classifier. Medical experts prefer text brightness or background color, and their view of the explanations’ usefulness varies with the visualization.

  7. 2025

    Hate Explained and social simulations

    In 2025, intent-tagged data improves the hate-speech classifiers in “Hate Explained,” and moderators using intent annotations outperform machine classifiers by 12.9%. He also coauthors a CENIA study that uses a GPT-4o-powered agent-based simulator to model online conversations and study misinformation and opinion bubbles.

  8. As of October 2026

    Researcher at CENIA

    His coauthored 2026 ICLR paper, “Rethinking Layer Relevance in Large Language Models Beyond Cosine Similarity,” proposes a more reliable way to identify important model layers and build smaller models that retain performance. CENIA lists him as an investigator.

Career

  1. Since 2024CENIAPostdoctoral researcher

Education

  1. 2022Pontificia Universidad Católica de ChileEngineering doctorate in computer science

From the biography.

  1. Google had the most people shaping AI every year from 2014 to 2025Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  2. The United States employs an estimated 48% of people shaping AIWe estimate where the people shaping AI work today, how universities, companies and research institutes split across countries, and how country shares have changed since the 1960s.
  3. 43% of living people shaping AI work outside their country of birthWe estimate that 43% of living people shaping AI work outside their country of birth (520 of 1,198 people with known birth and employment countries).
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionWe estimate where the people shaping AI studied and where their careers led. Massachusetts Institute of Technology leads with 467 former students, including 291 recorded doctorates.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Andrés Carvallo yet. Send us a free-licensed photo.

  1. Neural language models for text classification in evidence-based medicine arxiv.org. Checked 20 October 2026.
  2. Defense of Andrés Carvallo haivis.ing.uc.cl. Checked 20 October 2026.
  3. Andrés Carvallo – Instituto Milenio Fundamentos de los Datos imfd.cl. Checked 20 October 2026.
  4. Andrés Carvallo - ACL Anthology aclanthology.org. Checked 20 October 2026.
  5. Conoce los proyectos de investigación que desarrollarán los ganadores(as) del Fondecyt 2024 de Iniciación y Postdoctorado cenia.cl. Checked 20 October 2026.
  6. Unveiling Social Media Comments with a Novel Named Entity Recognition System for Identity Groups arxiv.org. Checked 20 October 2026.
  7. User Perception of Attention Visualizations: Effects on Interpretability Across Evidence-Based Medical Documents arxiv.org. Checked 20 October 2026.
  8. Investigadores CENIA crean “gemelo digital” de las conversaciones en redes sociales para estudiar discursos de odio y desinformación cenia.cl. Checked 20 October 2026.
  9. CENIA logra cuatro publicaciones en ICLR 2026, una de las conferencias más competitivas de IA a nivel mundial cenia.cl. Checked 20 October 2026.
  10. Investigación - CENIA cenia.cl. Checked 20 October 2026.