PyData Zurich: Causal Inference & Python Package Managment

After a long time of hibernation, the PyData Zurich meetup is finally back! We would like to rekindle the community and provide regular meetups with talks revolving around all things data in the Python ecosystem.

Schedule:

• 5:45 doors open
• 6:00 introduction
• 6:10 Talk #1 (30’ + 10’ questions)
• 6:50 Talk #2 (30’ + 10’ questions)
• 7:30 Apéro
• 9:00 Sending people home

Talk 1: Uplift Modelling: An Overview - Matteo Courthoud

Abstract: When evaluating a policy, we are interested in its incremental effect, the uplift. Uplift modelling combines techniques from causal inference and machine learning to estimate individual effects. Knowing the distribution of incremental effects rather than just the average allows for a better understanding of the policy mechanism and for effective targeting. In this talk, we will overview existing uplift modelling techniques from both a theoretical and applied perspective, using a toy business case.

Talk 2: Python Package Management: Future, Present, Past - Wolf Vollprecht

Abstract: Have you ever wondered how a package manager works? What are the differences between pip and conda and mamba? This talk gives a brief history of software package management software, particularly in the Python space, and shines some light on more recent advancements.

We will discuss

• what wheel files are, sdists and bdists, as well as conda packages.
• how semantic and calendar versioning schemes function
• whether you should use upper bounds to constrain dependencies or not!

Recently the team at prefix.dev took a deep dive into the Python resolver (resolvelib) and ported it to Rust – we will also discuss how that went, and how SAT solvers make modern package management possible. As an outlook we will take a look at some up-and-coming technologies in the package management space:

• rye, a new Rust based Python package manager
• pixi, a new Rust based conda / Python package manager
• sigstore, a modern way of signing packages, that replaces PGP

This event is kindly hosted by QuantCo. Find out more at https://www.quantco.com/.