×
Pietro Moriello

Pietro Moriello


Italian, English, Japanese

Background


Work Experience

Work Experience

  • Research Intern, Fujitsu Limited

    Oct, 2025 - Apr, 20266 months

    • Developed a proof-of-concept method for accelerating single-request decoding of Transformer language models quantized with Double Binary Factorization

    • Implemented and evaluated CUDA kernels for the proposed method; integrated them within PyTorch, enabling deployment with the HF Transformers library

  • Thesis Intern, University of Modena and Reggio Emilia

    Feb, 2025 - Aug, 20256 months

    • Conducted research on the adaptation of pre-trained neural networks for dynamic inference during the Master's thesis project

  • Software Developer, Studio Andrea Santachiara s.s.

    Sep, 2020 - Aug, 20265 years 11 months

    • Developed a C#/.NET GUI application and a library for the analysis of gas-chromatography measurements

    • Built an Android application to manage magnetic inductive sensors for the inspection of elevator ropes

Projects Experience

Projects Experience

  • Semantic Segmentation of Satellite Imagery

    - Present

    Computer Vision and Cognitive Systems course project

    • Trained and evaluated neural networks with PyTorch and HF Transformers for segmentation of satellite images on datasets with heavily imbalanced class distributions

    • Compared a retrieval method based on the Earth Mover's Distance of segmentation masks against cosine similarity on features from a DINO ViT encoder

  • WATERWISE

    - Present

    IoT and 3D Intelligent Systems course project

    • Deployed a tiny deep neural network on the ESP32 SoC using TensorFlow Lite to detect water leakage in plumbing systems

    • Implemented MQTT communication over WiFi to collect data from a distributed sensor network

    • Programmed the ESP32 firmware with the ESP library and RTOS framework

  • F1Trail

    - Present

    Real Time Embedded Systems course project

    • Developed a stereo vision system with OpenCV (C++) so an F1Tenth car can autonomously follow an ArUco marker

    • Integrated the system with ROS2 in C++

  • LoopBlocking

    - Present

    Educational project

    • Implemented a loop blocking optimization pass in LLVM (C++)

Skills

Skills

  • Programming Languages

    Python

    C/C++

    CUDA

    OpenMP

    C#

  • Tools and Frameworks

    PyTorch

    NumPy

    Pandas

    Git

    bash

    Linux

    Typst

Education

Education

  • Information and Communication Technologies (ICT), AImageLab, Ph.D., University of Modena and Reggio Emilia

    Nov, 2026 - Present

  • Artificial Intelligence Engineering, M.S., University of Modena and Reggio Emilia

    Oct, 2023 - Apr, 2026

    Thesis Supervisor: Prof. Simone Calderara

  • Computer Engineering, B.Eng., University of Modena and Reggio Emilia

    Sep, 2020 - Oct, 2023

    Final Project Supervisor: Prof. Andrea Marongiu

Publications

Publications

  • IrekoGPT: Turning Structured Pruning into Post-Hoc Slimmable LLMs , arXiv preprint arXiv:2610.00426

    Published on: Sep 30, 2026

    Pietro Moriello, Pietro Buzzega, Angelo Porrello, Simone Calderara

  • LUT-DBF: Leveraging Lookup Tables for Accelerating Quantized LLMs, 203rd High-Performance Computing Research Conference 2026, Sapporo, Japan

    Published on: Jan 01, 2026

    Pietro Moriello, Takashina Godai, Yusuke Nagasaka