Abstract
What does it look like to build an audio DSP pipeline in Rust?
In this talk, we will go from raw audio samples to frequency-domain features using a composable, high-performance pipeline. Along the way, we implement core primitives like windowing and FFTs, and explore how to structure processing stages without unnecessary allocations.
We’ll look at zero-copy data flow, trait-based design, and the tradeoffs between batch and real-time systems. The talk also includes performance comparisons with Python-based approaches, and practical lessons from optimizing real pipelines.
If you’re interested in systems programming, signal processing, or just making things fast without giving up safety, this talk is for you.
