Michele Ghittoni
Computer Science student and machine learning engineer focused on AI systems, experimental infrastructure, and product-minded software.
Michele Ghittoni is a Computer Science student at the Universita di Parma, graduating with a final grade of 110/110 cum laude.
His work spans machine learning experimentation, C++ systems for automatic differentiation, and full-stack product engineering. He has built research pipelines for HPC environments, co-founded a nightlife discovery platform, and independently shipped web products.
Education
- B.Sc. in Computer ScienceUniversita di Parma, Sep 2022 - Mar 2026
GPA 27.5/30. Final grade 110/110 cum laude.
Work
- Machine Learning Engineer InternUniversita di Parma, Oct 2025 - Feb 2026
Architected a scalable experimental framework to benchmark Dissimilarity Representation against traditional feature-based ML methods, with automated Python pipelines running on HPC through Papermill, Apptainer, and SLURM.
Skills
Machine Learning
PyTorch · Deep Learning · LLM · CUDA
Programming
Python · C · C++ · SQL · TypeScript
Frontend and Mobile
React · Next.js · SwiftUI
Systems and Infrastructure
HPC · AWS · Linux · Git
Projects
- Quickgrad2026
Open-source lightweight automatic differentiation and backpropagation engine built from scratch in C++ with a dynamic computational graph.
- Club-now2024 - 2026
Co-founded and led frontend engineering for a nightlife discovery platform with a React/TypeScript web app and native iOS app in Swift.
- Negozio-Codici2021
Independently developed and deployed an online platform for discovering discount codes.
- SerieA-News / TellDiff2018 - 2020
Built two social media pages to 35k and 25k followers through content strategy, analytics, and community management.
Languages
- ItalianNative
- EnglishFluent
- FrenchBasic
- SpanishBasic
Writing
- Building Quickgrad2026
Notes on building a small automatic differentiation engine from first principles.
- Dissimilarity Representation in Complex Geometric Settings2026
A short overview of an experimental framework for benchmarking dissimilarity-based ML methods.