The European Accessibility Act is now in effect, and you've done the hard work. Your app has passed its accessibility audit and complies with the latest legal standards. An Amazing milestone, but accessibility isn't a one-time fix! It's an ongoing commitment.

So, what happens next? How do you ensure your application stays accessible as your codebase evolves?

In this talk, we'll explore practical strategies for maintaining accessibility long after the initial audit. We'll look at how to embed accessibility into your development workflow, from automated tests to manual checks, and discuss where automation falls short. I’ll share my favorite tools and testing techniques, and explore practices that help prevent regression.

You'll leave with actionable insights to future-proof the accessibility of your application and ensure your update don't moves your application backward.

Jfokus

A conference talk transcript by Holden Karau (Spark contributor at Snowflake) covering Apache Spark performance optimization in lakehouse environments. The core problem: adding type information to Spark Datasets causes 10–400% slowdowns because Spark must validate full records instead of reading only needed columns. The solution being explored is transpilation — converting Python ASTs and JVM bytecode into Spark's Catalyst intermediate representation, allowing predicate and column pushdowns to work even with typed UDFs. A proof-of-concept exists with two passing tests, using recursive descent conversion of Python ASTs in Scala. Key challenges include silent correctness bugs (null handling, date/timezone semantics), avoiding broken pipelining when inserting select statements, and licensing issues with useful bytecode analysis libraries. Future work includes handling maps/flatMaps, opt-in relaxed typing, and targeting GPU operations (NVIDIA-style) beyond Catalyst.

Apache Spark Lakehouse Pushdowns with Transpilation, by Holden Karau