CV

Instructor — Summer School on Artificial Intelligence in Health and Life Sciences

Università Campus Bio-Medico di Roma

Short teaching appointment delivering lectures and hands-on sessions on neuroimaging-based brain age prediction.

• Delivered theoretical and practical sessions on brain age prediction from neuroimaging features using statistical, machine learning, and deep learning methods (Python; scikit-learn; statsmodels).
• Released a custom Kaggle dataset and accompanying code repository.

Researcher in Applied Artificial Intelligence

University of Tübingen

Research role in applied AI for neuroimaging and clinical/psychiatric datasets, focusing on explainable AI and generative modeling.

• Collaborating within the Mental Health Mapping Lab investigating 2D and 3D computer vision models through explainable AI (XAI) and generative model–based algorithms.
• Designing an end-to-end PyTorch-based ML workflow for large neuroimaging datasets to validate XAI explanations and probe model reasoning for mental disorders classification (CNNs) with Diffusion Models; using FreeSurfer, CAT, MONAI, batchgenerators, Lightning, Accelerator, W&B, MLflow; applying LLM tooling to psychiatric/clinical datasets; contributing to medical image segmentation foundation models (SAM).
• Handling large-scale imaging (DICOM/NIFTI) and tabular data (~70k images, ~300k records) with intensive validation, processing, and QA for downstream training and analysis.
• Managing a Slurm-based HPC cluster for distributed and parallel training of deep learning models.

DevOps Engineer, Telco Consultant

Hewlett Packard Enterprise (HPE)

DevOps/consulting role delivering CI/CD, backend integration, and delivery automation for a major Telco client.

• Led a development team for a major Telco client, managing continuous integration of feature requests and collaborating with customer stakeholders.
• Designed and implemented a Jenkins CI/CD pipeline (including automated SQL deployments) integrating with a Spring-based automation tool via API, reducing deployment time by ~90% and significantly cutting manual errors.
• Implemented new business logic and integration features in Java and SQL for backend services.
• Coordinated version control via Git across internal and client repositories and managed Dockerization of parts of the backend platform.
• Collaborated with client stakeholders to clarify requirements and translate business needs into technical designs and implementation plans.

IT Consultant

Concept Solution SRL

Consulting role focused on solution design and development planning across multiple emerging technology domains.

• Authored solution design and development plans for systems using user profiling, robotic process automation (RPA), blockchain-based traceability, and recommendation systems for e-commerce in close collaboration with business stakeholders.

Master

Politecnico di Milano

Computer Science and Engineering

• Artificial Intelligence focused path
• Thesis on Reinforcement Learning

Bachelor

Università degli Studi di Napoli Federico II

Computer Science and Engineering

• Computer science, mathematics, and physics background (incl. databases, software engineering, cybersecurity)

Medsamix: A training-free model merging approach for medical image segmentation

arXiv

arXiv preprint arXiv:2508.11032 (AAAI). Authors: Yang Y., Su G., Hu J., Sammarco F., Geiping J., Wolfers T.

Lightweight Model for Session-Based Recommender Systems with Seasonality Information in the Fashion Domain

ACM (Recommender Systems Challenge)

In Proceedings of the Recommender Systems Challenge. Authors include Sammarco F. et al.

Thesis: Non-Stationary Reinforcement Learning for Trading

• Built an automatic trading system for Forex assets using FQI reinforcement learning (gym).
• Integrated clustering and gradient-boosted classification (XGBoost) techniques.

Recommender System Challenges (RecSys, H&M, Internal)

• Developed a TV-series recommender using EASEr, matrix factorization methods, and SLIM.
• Built fashion-oriented recommenders using GRU4Rec and latent factorization approaches; trained via AWS.
• Participated in H&M and Dressipi challenges (Dressipi associated with RecSys Conference 2022).

Artificial Neural Network Challenges

• Built architectures for image segmentation/detection (incl. YOLO-style approaches), text classification, and sequence modeling (RNNs).

Website with Java supported back-end

• Implemented quiz logic, scoring, and user management features.

Thesis on Genetic Algorithms

• Developed a Python-based genetic algorithm system for job-shop scheduling.