1 of 11

  • ML = ?

Luca Palmieri

@algo_luca / LukeMathWalker

Engineer, TrueLayer

2 of 11

The slope of enlightenment

The ML industry is growing out of its infancy.

Businesses of every size are embedding ML in their products.

�Are we set up for success?

3 of 11

The performance framework

Accelerate, Nicole Forsgren PhD et al.

  1. Lead Time
  2. Deployment Frequency
  3. Mean Time to Restore (MTTR)
  4. Change Fail Percentage

4 of 11

ML 👑 - Python

General purpose, high-level

Easy to learn

Massive ML ecosystem

5 of 11

The ML Frontend

Little control over resources 😕

Poor performance 😕

...but great FFI! 😎

6 of 11

C/C++

Rust

The compiler knows�what I don’t know�🤝

I don’t know �what I don’t know�🔥

7 of 11

Rust impacts the bottom line

The compiler has cascading effects!

Safe route from WTF!? to production

Lower barriers to entry

Larger community

Healthier and fairer ecosystem

8 of 11

N-Dimensional Arrays

Feature-rich

Expanding ecosystem

Polishing-phase

ndarray

ndarray-stats

ndarray-linalg

ndarray-rand

ndarray-odeint

ndarray-vision

...

9 of 11

DataFrames

Design phase

Ongoing community discussion�https://github.com/rust-dataframe/discussion

?

Apache Arrow

10 of 11

Machine Learning

Several unmaintained crates

A lot of lurking interest

Let’s start talking! https://github.com/rust-ml/discussion

?

11 of 11

The plan

ndarray

ndarray ecosystem

${rust ml}

Apache Arrow

Python ecosystem

FFI

${rust dataframe}

Other data ecosystems