Biomedical AI · Explainable AI · Low-Data Learning · Biometrics · Compact Models · MLE-NITs

Athanasios (Thanos) Angelakis

Biomedical AI researcher with more than 10 years of production AI experience in the biomedical domain, building compact, explainable, low-data models for diagnosis, prognosis, biomarkers, medical imaging, and biometrics.

Consultancy, audits, and seminars available in English and Greek - online, onsite, or hybrid.

Postdoctoral Researcher, AI & Biometrics Security
BioML Lab, CODE Research Institute, Universität der Bundeswehr München, Munich, Germany

Affiliated Researcher / Sr. AI Data Scientist
Amsterdam UMC, Amsterdam, The Netherlands

Research methods Consultancy Education & Seminars GitHub Email

Core theme

Democratizing trustworthy biomedical AI

Compact, explainable, train-from-scratch AI for low-data medicine - when data are limited, labels are expensive, and massive GPU workflows are not realistic.

Low-data medicine

Useful AI for small cohorts, rare diseases, ultrasound, omics, biomarkers, and early diagnostic/prognostic tests.

CPU-friendly models

ZACH-ViT, hZACH-ViT, and s-DNNs: compact models designed to be trained, audited, and reproduced without massive compute.

Research profile

Trustworthy AI for biomedical and biometric high-stakes systems: tabular, imaging, omics, text, video, and multimodal data.

Method strengths

  • XAI, feature selection, low-data learning, multimodal AI
  • Model validation, bias/confounding, robustness, reproducibility
  • Biomarkers, diagnostic tests, prognostic models, clinical decision support

Methodological contributions

ZACH-ViT & hZACH-ViT

Compact vision transformers for low-data medical imaging, trained from scratch and stress-tested for robustness.

ZACH-ViT GitHub

s-DNNs

Small neural networks for tabular clinical data, non-invasive tests, biomarkers, and prognostic modeling.

MLE-NITs

Machine-learning-enhanced clinical scores and non-invasive tests, including MASLD fibrosis applications.

XAI feature selection

Feature-selection methodology for biomarker discovery, biological signal retrieval, and potential drug-target identification.

Research themes

Trustworthy ML Explainable AI Feature Selection s-DNNs MLE-NITs ZACH-ViT hZACH-ViT Non-Euclidean Geometric Deep Learning ShuffleStrides Data Augmentation Medical Imaging Biometrics ECG Authentication Privacy-Preserving Biometrics Robustness Tabular ML Omics Clinical Decision Support Liver Disease MASLD LLMs RAG Synthetic Data

Reviewer service

Reviewer for IEEE Access, IJCV, Knowledge-Based Systems, Journal of Hepatology, CSBJ, Nutrition & Metabolism, and JMIR.

Professional profile

I am a biomedical AI researcher and consultant working across academia, clinical research, and industry, with more than 10 years of production AI experience in the biomedical domain. My work focuses on compact, explainable, and auditable AI for high-stakes biomedical and biometric systems - from low-data medical imaging and biomarker discovery to diagnostic/prognostic tests, MLE-NITs, and privacy-aware biometric security.

Selected publications and submissions

hZACH-ViT: Curved Latent Geometry for Compact Vision Transformers in Low-Data Medical Imaging

A. Angelakis. arXiv, 2026.

arXiv:2606.00906

Machine-Learning-Enhanced Non-Invasive Testing for MASLD Fibrosis: Shallow-Deep Neural Networks Versus FIB-4, Tabular Foundation Models, and Large Language Models

A. Angelakis, G. De Vito, E.-M. Trifylli, F. Ferrucci. arXiv, 2026.

arXiv:2605.20523

Extending ZACH-ViT to Robust Medical Imaging: Corruption and Adversarial Stress Testing in Low-Data Regimes

A. Angelakis, M. Gomez Barrero. CVPR Workshops / PHAROS AIF-MIH Workshop, accepted, 2026.

arXiv:2604.06099

ZACH-ViT: Regime-Dependent Inductive Bias in Compact Vision Transformers for Medical Imaging

A. Angelakis. arXiv, 2026.

doi.org/10.48550/arXiv.2602.17929

ZACH-ViT: A Zero-Token Vision Transformer with ShuffleStrides Data Augmentation for Robust Lung Ultrasound Classification

A. Angelakis et al. arXiv, 2025.

doi.org/10.48550/arXiv.2510.17650

A selective machine learning algorithm for severe periodontitis labeling from questionnaire data

E. Stamatelou, N. Nijland, N. Su, B. G. Loos, A. Angelakis. Scientific Reports, 2026.

doi.org/10.1038/s41598-026-43934-6

LLMs for Drug-Drug Interaction Prediction Using Textual Drug Descriptors

G. De Vito, F. Ferrucci, A. Angelakis. Knowledge-Based Systems, 2026.

doi.org/10.1016/j.knosys.2026.115486

HELIOT: LLM-Based CDSS for Adverse Drug Reaction Management

G. De Vito, F. Ferrucci, A. Angelakis. Knowledge-Based Systems, 2025.

doi.org/10.1016/j.knosys.2025.114184

Extracellular vesicles as biomarkers for MASLD staging using explainable artificial intelligence

Trifylli EM, Angelakis A, et al. World Journal of Gastroenterology, 2025.

doi.org/10.3748/wjg.v31.i22.106937

Diagnosis of Acute Myeloid Leukaemia on Microarray Gene Expression Data Using Categorical Gradient Boosted Trees

A. Angelakis, I. Soulioti, M. Filippakis. Heliyon, 2023.

doi.org/10.1016/j.heliyon.2023.e20530

Adelic point groups of elliptic curves

A. Angelakis, P. Stevenhagen. Acta Arithmetica, 2021.

doi.org/10.4064/aa171025-27-3

Teaching, supervision, and education

Teaching

  • Adjunct Professor, Eindhoven University of Technology / JADS: Deep Learning; Data Structures and Algorithms.
  • Lecturer, University of Athens: Number Theory.
  • Teaching Assistant, Universiteit Leiden: Algebraic Number Theory and Algebra I.
  • Teaching Assistant, University of Athens: Matlab laboratories.

Supervision

  • PhD supervision in periodontitis diagnosis using proteomics and explainable AI.
  • PhD supervision in biometric recognition, face recognition, and fully homomorphic encryption.
  • Master student supervision across gene expression, OSM data quality, U-Net, speech recognition, forecasting, XAI, route optimization, radiomics, and demand forecasting.

Education

  • Ph.D. in Mathematics, Universiteit Leiden and Université de Bordeaux - Computational-Algorithmic Algebraic Number Theory.
  • M.Sc. in Applied Mathematics, National Technical University of Athens - Computational Mathematics.
  • B.Sc. in Mathematics, National and Kapodistrian University of Athens - Applied Mathematics / Computational Mathematics.

Recent invited talks

Education & Seminars

Responsible AI education for physicians and biomedical teams

I offer seminar and workshop formats for physicians, residents, clinical researchers, and pharmaceutical partners who need practical, clinically grounded education on responsible AI, multimodal biomedical AI, clinical decision support, and preparation for the agentic AI era in medicine.

For medical audiences

From FAIR clinical data and nonlinear diagnostic models to explainable AI, computer vision, LLMs, and safe clinical integration.

For pharma-supported education

Independent, scientifically rigorous AI seminars for collaborating MDs and clinical teams, adapted to oncology, hepatology, radiology, ultrasound, omics, or broader clinical AI literacy.

View seminar formats