CV

My professional and academic record. The complete CV is available for download in PDF format by clicking the PDF icon.

General Information

Full Name Daniel DeAlcala
Date of Birth 6th August 1998, Barcelona, Spain
Phone (+34) 690918474
Email danieldealcalavalcarcel@gmail.com
LinkedIn daniel-dealcala

Education

  • Jan 2022 – expected early 2027
    Ph.D. in Telecommunications and Computer Engineering
    Universidad Autónoma de Madrid (UAM)
    • Thesis: Auditing Artificial Intelligence: Membership Inference and Methods for Transparent, Fair and Trustworthy AI
    • Advisors: Prof. Julian Fierrez and Prof. Aythami Morales
    • Funded by an FPU fellowship (FPU21/05785), Spain's national 4-year predoctoral fellowship
  • Sept 2020 – Sept 2021
    M.Sc. in Deep Learning for Audio and Video Signal Processing
    Universidad Autónoma de Madrid (UAM)
    • Thesis: egocentric video localization and multi-user graph modelling for contact tracing with life-logging cameras
  • Sept 2016 – Jun 2020
    B.Sc. in Telecommunications Engineering
    Universidad Autónoma de Madrid (UAM)
    • Ranked 1st in the graduating class
    • 8 Matrículas de Honor (top distinction, capped at 5% of each course)
    • UAM Excellence Scholarship; EPS Effort Award
    • Thesis: Egocentric Biometric Palmprint Recognition for VR

Research Experience

  • Sep 2026 – Dec 2026
    Visiting Researcher, Docling team (host Peter Staar)
    IBM Research – Zurich, AI for Knowledge. Zurich, Switzerland
    • Contributor to Docling, IBM's open-source document conversion toolkit (60K+ GitHub stars), working on several research projects around it
    • Developing a VLM that links the fields of a form to their values
  • Sep 2025 – Dec 2025
    Visiting Researcher (host Pin-Yu Chen)
    IBM Research, MIT-IBM Watson AI Lab. New York, USA
    • Took AI-generated content detection from single images or text passages to whole documents. Built AIdocFactory, an open-source pipeline that parses PDFs with Docling, rewrites chosen paragraphs, figures and tables with any Hugging Face generative model, re-renders the page and records box-level labels. Used it to build a 35K-page benchmark from 2,000 arXiv papers
    • Trained AIdocDetector, a 256M-parameter VLM that classifies each page and segments the edited regions: F1 of 0.83 on seen and unseen generators and Dice of 0.85–0.88, above GPT-5.2, Gemini 3 Flash and Claude Opus 4.5 prompted as detectors
    • First-author paper in submission, journal extension in preparation; IBM filed a patent application on the method
  • Jan 2022 – present
    PhD Researcher and FPU Fellow
    BiometricsAI Lab (formerly BiDA Lab), Universidad Autónoma de Madrid
    • DeepID Challenge at ICCV 2025 (forged ID documents): designed and trained the UAM entry, TruFor fine-tuned with a SegFormer backbone and knowledge distillation. 1st of 26 teams on the private test set of 20K real ID documents, in both tracks, and 2nd overall in manipulation localization
    • Introduced the Membership Inference Test (MINT): activation-based MINT reaches up to 90% accuracy on 22M+ face images (IEEE Access 2025); Active MINT stays above 80% from MobileNet to ViT (ICCV 2025); gMINT reaches near-perfect detection in most settings (CVPR-W 2025)
    • Co-developed MINT for language models: 7 Transformer models and 2.5M sentences, AUC 0.85–0.99 (IEEE CAI 2026), and LoRA-fine-tuned LLMs, precision 0.77–0.92 across 4 models (preprint, 2026)
    • Built ai-mintest.org, a public web tool that runs these audits on image and text models (AAAI 2025 workshop demo; IEEE COMPSAC 2026)
    • Other first-author work: AttZoom, a drop-in spatial attention layer for CNNs (ICCV-W 2025); N-Sigma, a statistical framework for measuring bias in AI models (IEEE COMPSAC 2023)
  • 2026 – present
    ReTeX, Independent Research Project
    Personal project
    • Training a vision-language model that turns images of document pages into LaTeX source, treating document conversion as image-to-code generation. Paper in preparation. Project page

Awards

  • 2025
    Best score on the private test set (20K real ID documents), DeepID Challenge
    ICCV 2025
  • 2022
    FPU Fellowship (FPU21/05785)
    Spanish Ministry of Universities
  • 2020
    Ranked 1st in class, UAM Excellence Scholarship, EPS Effort Award
    Universidad Autónoma de Madrid

R&D Projects

  • BBforTAI through MICINN/FEDER under Grant PID2021-127641OB-I00
  • HumanCAIC through MICINN under Grant TED2021-131787B-I00
  • M2RAI through MICIU/FEDER under Grant PID2024-160053OB-I00
  • Cátedra ENIA UAM-VERIDAS (NextGenerationEU) under Grant TSI-100927-2023-2

Service and Teaching

  • Reviewer
    • NeurIPS, ICML, ICLR, CVPR and ICCV (2023 – 2026)
    • AAAI workshops and IEEE Transactions on Biometrics, Behavior, and Identity Science (T-BIOM)
  • Teaching
    • Teaching assistant, Multimedia Signal Processing, B.Sc. in Telecommunications Engineering, UAM (2023 – present)

Publications

  • [1] D. DeAlcala, A. Morales, J. Fierrez, G. Mancera, R. Tolosana, J. Ortega-Garcia. Active Membership Inference Test (aMINT): Enhancing Model Auditability with Multi-Task Learning. ICCV 2025.
  • [2] D. DeAlcala, A. Morales, J. Fierrez, G. Mancera, R. Tolosana, J. Ortega-Garcia. Is My Data in Your AI? Membership Inference Test (MINT) Applied to Face Biometrics. IEEE Access, 2025.
  • [3] P. Korshunov, ..., D. DeAlcala, ... (34 authors). DeepID Challenge of Detecting Synthetic Manipulations in ID Documents. ICCV 2025.
  • [4] G. Mancera, D. DeAlcala, J. Fierrez, R. Tolosana, A. Morales. Is My Text in Your AI Model? Gradient-based Membership Inference Test Applied to LLMs. IEEE CAI 2026.
  • [5] D. DeAlcala*, G. Mancera*, J. Fierrez, A. Morales, R. Tolosana, R. Vera-Rodriguez. Is My Vision-Language Data in Your AI? Membership Inference Test (MINT) Demo 2. IEEE COMPSAC 2026.
  • [6] D. DeAlcala, I. Serna, A. Morales, J. Fierrez, J. Ortega-Garcia. Measuring Bias in AI Models: An Statistical Approach Introducing N-Sigma. IEEE COMPSAC 2023.
  • [7] D. DeAlcala, C.-Y. Ko, A. Morales, J. Fierrez, R. Tolosana, P.-Y. Chen. DocAI: Detecting and Localizing AI-Edited Content in Documents. Under review at ICLR 2027.
  • [8] G. Mancera, D. DeAlcala, A. Morales, J. Fierrez, R. Tolosana, F. Jurado. Auditing Training Data in Domain-adapted LLMs: LoRA-MINT. arXiv:2606.06946, 2026.
  • [9] M. Lopez-Duran, E. Marrero, J. Fierrez, M. Robledo-Moreno, R. Vera-Rodriguez, D. DeAlcala, A. Morales, R. Tolosana, O. Delgado, A. Ortigosa, J. Ortega-Garcia. Comparative Study of Domain-adapted VLMs for General Document Visual Question Answering. ICDAR 2026 Workshops.
  • [10] M. Lopez-Duran, J. Fierrez, A. Morales, D. DeAlcala, G. Mancera, J. Irigoyen, R. Tolosana, O. Delgado, F. Jurado, A. Ortigosa. CrimeNER Demo: Named-Entity Recognition in the Crime Domain. ICDAR 2026 Workshops.
  • [11] D. DeAlcala, A. Morales, J. Fierrez, R. Tolosana. AttZoom: Attention Zoom for Better Visual Features. ICCV 2025 Workshops.
  • [12] D. DeAlcala, A. Morales, J. Fierrez, G. Mancera, R. Tolosana. gMINT: Gradient-based Membership Inference Test Applied to Image Models. CVPR 2025 Workshops.
  • [13] D. DeAlcala, A. Morales, J. Fierrez, G. Mancera, R. Tolosana, R. Vera-Rodriguez. MINT-Demo: Membership Inference Test Demonstrator. AAAI 2025 Workshop on AI Governance.
  • [14] G. Mancera, D. DeAlcala, A. Morales, R. Tolosana, J. Fierrez. Membership Inference Test: Auditing Training Data in Object Classification Models. AAAI 2025 Workshop on Deployable AI.
  • [15] D. DeAlcala, G. Mancera, A. Morales, J. Fierrez, R. Tolosana, J. Ortega-Garcia. A Comprehensive Analysis of Factors Impacting Membership Inference. CVPR 2024 Workshops.
  • [16] D. DeAlcala, A. Morales, R. Tolosana, A. Acien, J. Fierrez, S. Hernandez, M. A. Ferrer, M. Diaz. BeCAPTCHA-Type: Biometric Keystroke Data Generation for Improved Bot Detection. CVPR 2023 Workshops.
  • [17] I. Serna, D. DeAlcala, A. Morales, J. Fierrez, J. Ortega-Garcia. IFBiD: Inference-Free Bias Detection. AAAI 2022 Workshop on AI Safety (SafeAI).
  • [18] R. Daza, D. DeAlcala, A. Morales, R. Tolosana, R. Cobos, J. Fierrez. ALEBk: Feasibility Study of Attention Level Estimation via Blink Detection Applied to e-Learning. AAAI 2022 Workshop on AI for Education.

Conference Participation

  • Computer Vision and Pattern Recognition Conference (CVPR), 2023 - Vancouver, Canada
  • Computers, Software, and Applications Conference (COMPSAC), 2023 - Turin, Italy
  • Computer Vision and Pattern Recognition Conference (CVPR), 2024 - Seattle, WA, USA
  • The Association for the Advancement of Artificial Intelligence (AAAI), 2025 - Philadelphia, Pennsylvania, USA
  • Computer Vision and Pattern Recognition Conference (CVPR), 2025 - Nashville, TN, USA
  • International Conference on Computer Vision (ICCV), 2025 - Honolulu, Hawaii