Federico Rossi

Ultra low-latency connectivity for foreign exchange market access.

Software engineering and formal verification. Formalization, implementation, and maintenance of a PyTorch-based framework for neural network validation, formal verification, and model checking.

Design, implementation, and verification of a RISC-V instruction set extension for ML model optimization. Development and maintenance of the cppPosit C++23 library. Porting the cppPosit library into tinyDNN and Tensorflow DNN frameworks.

Leader of Work package 2.1 for "Accelerators with Mixed-Precision". Development of IPs for numerical data compression and posit arithmetic in RISC-V cores.

Deployment of real-time man-down tracking algorithm for RISC-V and edge computing devices.

Thesis: "Innovative Arithmetics for Efficient DNN Computing, HW and SW Solutions and Their Integration in RISC-V Platforms"

Thesis: "On the Use of Tabulated Posits in Deep Neural Networks for Real-Time Image Classification"