Geographical Data & Symbolic Differentiation with Python

Hello PyData People!
We are excited to announce our next event of 2026! This time, we will be hosted at FlixBus Italia's headquarters in Milan for an evening dedicated to X
📅 When: Tuesday, June 23rd, 2026 – 18:30–21:00
📍 Where: FlixBus Italia, Corso Como 11, 20154 Milan
⚠️ Important: Spots are limited. Please keep your RSVP updated to allow others to participate if you can no longer attend.
***
🕒 Agenda

• 18:30: 🚪 Doors open & Check-in
• 19:00: 🎤 Talk 1: Process geographical data in Python – Jacopo Farina
• 19:45: 🎤 Talk 2: Symbolic Differentiation of Matrices and N-Dimensional Arrays in SymPy – Francesco Bonazzi
• 20:30: 🍕 Networking & Social Dinner

***
🎤 The Talks
1️⃣ Process geographical data in Python
Speaker: Jacopo Farina (Staff Data Engineer @ FlixBus)
In this talk, we are going to see how to process data from OpenStreetMap, public transport data (like the schedule from ATM) and other sources to calculate travel distances in a city, find addresses, find where and when people are using bike sharing (BikeMI) and nice things that can be done with this data. We are also going to see what OSM is and how it represents geographical data, how this data can be imported and used in Python (using PostGIS or DuckDB), how to process it and how to display the results.
About the Speaker:
Jacopo Farina is a data engineer at Flixbus, where he works on the demand forecasting, and a teacher at the Data Science Retreat in Berlin. Previously he worked on the backend using Node.js, reporting and ETL.
***
2️⃣ Symbolic Differentiation of Matrices and N-Dimensional Arrays in SymPy
Speaker: Francesco Bonazzi (PhD in Physics and Software Developer @ Intesa Sanpaolo)
Numerical frameworks such as PyTorch and JAX routinely compute matrix and tensor gradients via reverse-mode automatic differentiation (backpropagation) to train neural networks. By contrast, there has been a notable absence of symbolic implementations that produce exact, closed-form gradient expressions interpretable and well-suited to further algebraic manipulation.
SymPy, an open-source computer algebra system written in Python, now provides comprehensive support for symbolic differentiation of matrices and N-dimensional arrays.
The implementation gives users flexibility to work with either matrices or arrays, which can be handled in two ways: explicitly, by defining each component individually, or implicitly, as symbolic expressions.
In the explicit case, the derivation algorithm is straightforward: the derivatives essentially amount to nested loops over the array dimensions. By contrast, the implicit algorithm is considerably more intricate, especially when deriving expressions that involve matrices.
After outlining the key aspects of this algorithmic design, the talk showcases its practical value through selected examples that illustrate how closed-form symbolic matrix and tensor gradients can provide valuable insights and enable exact computations in various applications.
About the Speaker:
Francesco Bonazzi is a software developer and core maintainer of SymPy, a Python computer algebra system. He has contributed extensively to the library's physics, matrix algebra, tensor algebra, and probability modules. Currently, he works at Intesa Sanpaolo, focusing on the research and implementation of generative AI. Francesco holds a PhD in Physics and previously served as a researcher at the Max Planck Institute for Colloids and Interfaces in Germany and North Carolina State University in the U.S.
***
See you there! 🚀
The PyData Milano Team