What success looks like
- Routing choices rest on measured results, not on a feeling about a model.
- Each public study states what it did not test.
- A cheaper route ships only when the checks say it keeps the quality.
What you will do
- Design tasks, graders and protocols, and declare them before the first run.
- Run studies across models and harnesses, with intervals and honest tie calls.
- Turn a finding into a product change, then measure the change.
- Write the study in plain words for people who will check your work.
About you
- You know the difference between a result and a story about a result.
- You are comfortable with confidence intervals and with saying “we do not know yet”.
- You write clearly and you do not oversell.
Nice to have
- Experiment design
- Cost and latency analysis
- Public technical writing
A short note, please
With your resume, write a page or less on the work you are proudest of. Tell us what you chose to cut, and which of our values you would hold us to.