About the role
You turn promising ideas into experiments other people can trust and extend, working in the space between scientific exploration and production engineering. Most research code answers a question once and is then unreadable; yours has to answer it in a way somebody else can check and build on. Formal research experience is welcome here, and serious independent or open-source work counts just as much.
The work
Implement methods, reproduce papers, construct datasets and evaluation environments, and optimize experimental systems. Work with scientists to identify confounders and with product engineers to test transfer into real workflows. Document failures as carefully as successes.
What good looks like
In your first 90 days, reproduce one important baseline and deliver a well-tested experiment or prototype that answers a specific question.
Evidence we look for
Bring strong programming, quantitative reasoning and curiosity. Research experience is welcome, and equivalent evidence from serious independent or open-source work is equally useful.
What we need to see
- Strong programming, and quantitative reasoning you can show your working for
- Genuine curiosity, evidenced by something you built or investigated because you wanted to know
- You write experiments other people can run and extend
- You can judge whether a result is real before anybody else has to
Nice to have
- Research experience, formal or otherwise
- Open-source work anyone can read
- Depth in an area adjacent to our questions
The exercise
Reconstruct an experiment from incomplete notes and identify what must be clarified before its result can be believed.
Where and how we work
In the office together five days a week, in any of these cities. Remote-friendly around your family, arranged one person at a time.