Valerio Siniscalco
contact@siniscalco.me codeberg linkedin orcid scholarAbout
I am doing an MSc in IT and Cognition at the University of Copenhagen and work part-time as a student AI engineer at Retriever, where I fine-tune language models for media intelligence.
From January to July 2026 I was a research intern at the Centre for Mind/Brain Sciences in Trento with Prof. Uri Hasson, where I trained topographic convolutional networks and studied how the topography changes their representations and their robustness. I took my BSc in Computer Science at the University of Milan, and my thesis at ITU Copenhagen’s NERDS Lab with Prof. Michele Coscia is the case study of a CIKM 2026 paper on network backboning.
Publications
- Multilayer Network Backboning It extends network backboning to multilayer networks, with the Wikipedia network from my BSc thesis as the main case study.
Projects
- Topographic Neural Ensembles Code for the CIMeC ensemble experiments, in which ResNets trained at several strengths of a topographic regulariser are combined into ensembles and compared through their representations. Each experiment is a Hydra config logged to MLflow, with the models stored in S3, and it runs locally or on a cloud GPU.
- Wikipedia Network Backboning The pipeline and dataset from my BSc thesis, a multilayer network of 94,403 people across five Wikipedia editions, with 12.4 million candidate edges from entity linking, filtered by Noise-Corrected backboning against a statistical null model.
Timeline
-
Aug 2026 –
MSc, IT and Cognition current University of Copenhagen, Faculty of Humanities Machine learning, natural language processing, and computer vision, alongside human cognition and language.
-
Oct 2026 –
Student AI Engineer, part-time current Retriever In the AI Solutions team I fine-tune language models for media intelligence, and work on retrieval-augmented generation, model evaluation, and on-prem inference.
-
Sep 2026 –
Mentee, LeadTheFuture current LeadTheFuture is a mentorship non-profit for Italian students in STEM, with an acceptance rate below 20%. I was paired with a mentor one to one.
-
Jan – Jul 2026
Research Intern Centre for Mind/Brain Sciences (CIMeC), University of Trento, supervised by Prof. Uri Hasson I trained topographic convolutional networks in PyTorch, as ensembles of fully trained ResNets and as adapters on a frozen pretrained network. I measured how the topography changed their representations, with representational similarity analysis and measures of spatial organisation, and how robust they were to noisy inputs and perturbed weights. I set up the group’s experiment tracking and model storage with MLflow and S3, used by the team, and developed an open-source library for typed stage caching in MLflow.
-
Sep – Dec 2025
Research Intern Networks, Data, and Society (NERDS) Lab, IT University of Copenhagen, supervised by Prof. Michele Coscia BSc thesis, with a travel grant from the Danish Data Science Academy, on how connections between historical figures in five Wikipedia editions differ by gender, origin, and era. I found the connections with NLP entity linking and kept the statistically significant ones with Noise-Corrected backboning, a null-model test.
-
2025 – 2026
Co-Founder, Polyneta Smart-cities startup, Milan The startup won USC Milan with prototypes of privacy-preserving machine learning and NLP for municipal data.
-
2022 – 2026
BSc, Computer Science Università degli Studi di Milano 110/110 cum laude. Thesis: Representational Bias on Wikipedia: A Multilingual, Multi-Attribute Network Analysis, written at ITU Copenhagen’s NERDS Lab and supervised by Prof. Elena Casiraghi and Prof. Michele Coscia.
Music and books I like
- Time Out, The Dave Brubeck Quartet
- 30 νυχτερινά, Μάνος Χατζιδάκις
- L’isola del giorno prima, Umberto Eco
- La gatta Cenerentola, Nuova Compagnia di Canto Popolare
- Quaglia Sovversiva, Marco Castello
- Three Men in a Boat, Jerome K. Jerome
- Tino’s Breaks, Volumes 1 & 2, Tino
- Some Prefer Nettles, Jun’ichirō Tanizaki
- Samba esquema novo, Jorge Ben