Go directly to page contentent
TEAM LADE
Michele Alessi
Researcher at Laboratory of Data Engineering

I am a PhD student in Applied Data Science and Artificial Intelligence at the University of Trieste and Area Science Park. My research combines probabilistic modelling, representation learning, and single-cell genomics, with a focus on deep generative models for transcript-level data. I develop computational methods to study isoform usage and integrate biological signals across different levels of resolution, connecting artificial intelligence, data science, bioinformatics, and omics research.

Research Interests
  • Probabilistic and deep generative modelling for single-cell transcriptomics
  • Transcript- and isoform-level analysis, including differential transcript usage in tumour data
  • Integration of gene-level, transcript-level, and multimodal omics data
  • Variational inference and interpretable representation learning
  • Density-informed and clustering-based priors for latent-variable models
Experience & Education
  • 2024–present — PhD in Applied Data Science and Artificial Intelligence, University of Trieste and Area Science Park
  • 2022–2024 — MSc in Data Science and Scientific Computing, University of Trieste, University of Udine, SISSA, and ICTP; graduated with honours
  • 2019–2022 — BSc in Mathematics, University of Trieste; graduated with honours
  • 2025–present — Teaching Assistant in Deep Learning, University of Trieste
  • 2025 — Presenter at EurIPS, Copenhagen, and participant in international schools on generative and foundation models
  • 2024 — Machine Learning Engineer Intern, Asimov; worked on large language model fine-tuning and feature-masking methods
  • 2019–2024 — Merit Scholarship Recipient, Collegio Universitario Luciano Fonda
Latest Publications
03/12/2025
Density-Informed VAE (DiVAE): Reliable Log-Prior Probability via Density Alignment Regularization
Abstract We introduce Density-Informed VAE (DiVAE), a lightweight, data-driven regularizer that aligns the VAE log-prior…
Go to the news Density-Informed VAE (DiVAE): Reliable Log-Prior Probability via Density Alignment Regularization