Writing the Fastest GBDT Library in Rust by Isabella Tromba In this talk, I will share my experience optimizing a Rust implementation of the Gradient Boosted Decision Tree machine learning algorithm. With code snippets, stack traces, and benchmarks, we’ll explore how rayon, perf, cargo-asm, compiler intrinsics, and unsafe rust were used to write a GBDT library that trains faster than similar libraries written in C/C++.