Alessandro Cheli

Development of professional audio software in C++23 and Faust, with research of neural networks for realtime audio in Python. Inspired from modular and analog synthesizers. Commercial release targeted for Q4 2026.

Building and maintaining the industry standard React and Svelte libraries for building interactive node-based UIs for graphs and workflows: React Flow and Svelte Flow. Full stack integrations, from designing UI component systems, to maintaining the core library, improving and modernizing the DX and complex build systems of our internal repositories, to integrating external systems and solutions such as collaborative CRDT-based multiplayer libraries. All for a library with 5M weekly downloads.

Symbolic AI, applied category theory, compilers and programming languages. Building the core since the early stages of the company, and applying category theory to representations of reasoning and computation and interdisciplinary problems in cutting-edge explainable AI. Contributing to UI and infra tasks.

During GSoC 2024, I mentored Niklas Heim’s project OptiFloat.jl. Based on my project Metatheory.jl, OptiFloat.jl rewrites floating point expressions to more accurate alternatives. It is a pure Julia implementation of Herbie. To learn more about what OptiFloat does, how to use it, and how it works check out the documentation.

I completed my undergraduate studies in Computer Science at the University of Pisa. My thesis proposes an advanced, generic and high-level code rewriting and analysis system for the Julia programming language, and was supervised by Professors Gian-Luigi Ferrari, Andrea Corradini and Dr. Christopher Rackauckas

During GSoC 2021, I was awarded a stipend from Google for an experiment involving the integration of Metatheory.jl, my thesis project, with the high performance Computer Algebra System (CAS) Symbolics.jl developed at MIT. The project focused on automatically applying user-defined high-level compiler optimizations, defined via equational rewrite rules, to preemptively choose the symbolic representation of a system of differential equations that performs best when compiled and solved numerically. The projects are still being maintained and there is still room for exploring and researching the topic.

High school sponsored internship, focused on the development of full stack backoffice applications in Node.js and React.

Short intership, sponsored by the Erasmus program for high school students. I focused on development of various visual detection systems for embedded computers with OpenCV running on Raspberry Pis, and a mobile application in React Native.