Performance critical ML: How viable is Rust as an alternative to C++

Hey everyone! I have a small team that creates production machine learning algorithms and integrates them with apps.

They’re usually served with Python/FastAPI, but we’ve also written inference to run locally in iOS/WASM.

We do a fair amount of novel algorithm development and generally write the computationally intensive parts in C++. This is especially important for inference when milliseconds matter

Recently we’ve built some prototypes in Rust and it’s been amazing. Similar performance, but much nicer to write.

We’re considering adopting Rust as our default. I have a decent amount of C++ experience, but very little Rust.

Are there any issues/challenges I should be aware of before going too far down this road?