Abstract: With the many computational challenges that face Run 3 of the Large Hadron Collider (LHC), there is an increased demand to exploit the recorded data as much as possible. There is a complementary desire to empower analysts with toolkits that allow for expressive manipulation of the data and rapid iteration from analysis design to final insights.

In this talk, I will motivate the importance of differentiable programming in analysis toolkits and illustrate the opportunities for physics analyses at the LHC to be optimized from end to end by using efficient gradient-based optimization algorithms. I will additionally explore ongoing dedicated efforts across the Institute for Research and Innovation in Software for High Energy Physics (IRIS-HEP) to integrate differentiable programming techniques with highly scalable workflows to create performant end-user analysis tools, with applications beyond the scope of experimental high energy physics.