Abstract
Automatic differentiation is transforming scientific computing, enabling gradient-based approaches to optimization. It powers modern artificial intelligence, and now also transforms the world of scientific computing, enabling end-to-end workflows bridging both worlds. This talk introduces how JAX's composable transformations and the Equinox ecosystem make this practical: Equinox provides module system built on JAX's functional paradigm with Pytorch-like syntax, while libraries such as Diffrax for differential equations, Lineax for linear solvers, and Optimistix for nonlinear optimization and root-finding — offer performant numerics out of the box. We'll walk through key concepts and real examples showing how this ecosystem enables end-to-end differentiable scientific workflows.
 

Biosketch
Johanna is a maintainer of Optimistix and an active contributor to scientific computing libraries in the Equinox ecosystem in JAX. She is currently a PhD researcher in systems biology at ETH Zurich.