DataData Scientist

Find out which apps assistants write well, which they get wrong, and what we should tell them so the next one works on the first try.

All open roles
Function
Data
Region
Remote
Type
Full-time
Experience
3+ years
Apply for this role

About the role

Every app here is written by someone's assistant, and every version is kept with the checks it passed or failed and the warnings it carried. That's a growing record of what assistants actually produce when asked for a small personal app.

Your job is to learn from it: which requests fail the checks, which updates get rolled back, which assistants do better, and which changes to the instructions we give them make the first version work more often.

What you'd do

  • Study the apps and versions we hold to find the common ways an assistant's app fails
  • Propose and test changes to what the app MCP tells assistants, and measure the effect
  • Help tune the checks: what they refuse that's fine, and what they miss
  • Report plainly what the evidence supports and what it doesn't

What we need

  • Three or more years of applied data science or evaluation work
  • You've designed an experiment or evaluation and can explain why you trusted its result
  • Comfortable reading code, since the data here is mostly code

Nice to have, not required

  • Evaluating LLM output, especially generated code
  • HTML and JavaScript well enough to tell why an app breaks on a phone

How we work

  • Small team, thin slices, shipped weekly. Nothing sits on a branch for a month.
  • Reviews are real. Every feature ships with its tests, and a bug found in review is cheaper than one found by someone using it.
  • Private by default. A person's app data stays on their phone, and nothing we build may need to read it.
  • We only claim what ships. That applies to the product, the roadmap and the offer letter.

How hiring works

  1. 1A 30-minute call about something you've built and what was hard about it.
  2. 2A working session on a real problem from this codebase. You can drive, or we can pair.
  3. 3A conversation about how you make decisions when the evidence is thin.
  4. 4References, then an offer. We aim to answer within a week at every stage.

Apply: Data Scientist