Research engineer, evaluation
Build the tests that tell us when the model is better, and when it only looks better.
Apply for this role
C$170k–C$240k + equity
Good evaluation is how we keep ourselves honest. You will own the harness every checkpoint runs through.
What you’ll do
Build held-out sets that resist contamination
Automate evaluation across thousands of GPUs
Publish our methods so others can report the same numbers
You write careful code and are suspicious of numbers that look too good.
Open roles