Reproducible machine learning
You are a generalist engineer, leaning towards backend/infrastructure. You will basically help us build this thing from scratch. We don't mind what particular skills you already have. We figure you can pick up something new quickly.
When starting this company, we thought: instead of getting a job at the best place to work, let's make that best place to work. We want to work with the best people in an inclusive, supportive environment. And, just have fun while we're at it. You will help us make that place.
We're looking for the right person, not just someone who checks boxes, so you don't need to satisfy all these things. But, you might have some of these qualities:
We want our team to feel invested in what we're building, so we will give the right person a meaningful amount of equity. And, all the usual things. (We're European so you'll get really good healthcare.)
Machine learning research is really similar to open source software: it's a worldwide community of people sharing new techniques and building upon each other's work. But, the standard way of sharing new work is publishing a PDF on a mailing list from the 90s.
Using machine learning inside companies it isn't much better — there's no version control, random files are scattered on S3, nobody knows what's running where.
To fix this, we're building a place where researchers can publish an actual, runnable version of their work. Other researchers can run their work and build upon it. It will be a community of all the best ML researchers.
Inside companies, this will be the place where researchers and engineers store models. Models will be stored in one place in a runnable format, so other people can try them out, build upon them, and deploy them to production.
We're also building the open source infrastructure to make this work: tools for versioning and packaging machine learning models.
Ultimately, we want machine learning to be as collaborative and accessible as open source software. You shouldn’t have to understand complex math or write an academic PDF to do useful things with ML. It should be as easy as forking code on GitHub or importing a package from npm.