Speakers: Ralf Gommers (Quansight)

Abstract:
Modern scientific discovery relies heavily on foundational open-source libraries like NumPy and SciPy. While these tools underpin global research, they face a persistent "foundation tax". Simultaneously, the rapid expansion of AI infrastructure has created a fragmented computing landscape, where specialized hardware optimizations often require abandoning classic scientific Python workflows.
A new dynamic is emerging through ongoing ecosystem-wide collaboration, notably around the evolving Array API Standard. Rather than deep learning frameworks simply building on top of foundational Python, core scientific software is beginning to leverage the immense hardware investments poured into AI compute backends. By decoupling high-level scientific syntax from low-level execution, the ecosystem is moving toward a future where domain logic can run seamlessly across CPUs, GPUs, and other accelerators without rewriting the core codebase.
This seminar will explore this technical and cultural evolution. We will feature a performance demonstration comparing traditional native compiled workflows with modern Python code employing SciPy and JAX, showing how pure Python can match native speed on the CPU and scale across hardware. Alongside these performance metrics and the underlying infrastructure that enables them, we will examine the changing human dynamics of open source, including the current strain placed on maintainers by the influx of AI-generated code, and look ahead at sustainable organizational models for the digital commons.
Speaker's Bio:
 CEO Technology at Quansight, NumPy & SciPy Maintainer, Open Source Strategist. For over 15 years, Ralf has been a key contributor to the scientific Python, PyData, and Python packaging ecosystems. He maintains NumPy, SciPy, and meson-python, and is the creator of the Python Array API Standard and pypackaging-native. At Quansight, he drives open source innovation, fostering sustainable communities and advancing data science & scientific computing technologies.
This seminar is part of the CERN STEAM Academy Seminar Series. 
Networking cocktail will follow the seminar. 

A public webcast will be available and accessible to external participants.
With the support of CERN's Next Generation Triggers Project.

https://indico.cern.ch/event/1696050/

Zoom: https://cern.zoom.us/j/65314260698?pwd=N2iB5wTo5cyW5ZxqtO4xWCCzQYnMjj.1