Anthropic interviews
Anthropic interviewsResearch and engineering
How the process runs
“One thing worth knowing: engineers here do lots of research, and researchers do lots of engineering.”
Anthropic CareersCareers page, "What we are looking for"
Engineers do research, researchers engineer
Practise spending a fixed compute budget on the experiment that answers the question.
ExampleResearch judgment
“An area of human comparative advantage, for now, is research taste and judgment, including choosing which problems matter, which results to trust, and when an approach is a dead end.”
The Anthropic InstituteInstitute essay "When AI builds itself"
Research taste stays with people
Practise checking what a benchmark gain still supports after an overlap check.
Example“Similarly, once Claude can run experiments, the question shifts towards “Which of these experiments is worth running?”
The Anthropic InstituteInstitute essay "When AI builds itself"
Which experiments are worth running
Practise checking what a benchmark gain still supports after an overlap check.
Example“We find that hill-climbing on only one benchmark does not lead to a generalizable result.”
Chen Yueh-Han, Jiaxin Wen, Jan Hendrik Kirchner (Anthropic Alignment Science Blog)Research post, work done during the Anthropic Fellows Program
One benchmark is not enough
Practise checking what a benchmark gain still supports after an overlap check.
Example“The held-out benchmark tests generalization: it is never shown to the AAR”
Chen Yueh-Han, Jiaxin Wen and Jan Hendrik KirchnerPaper coauthors, work done during the Anthropic Fellows Program
Hold out a benchmark
Practise checking what a benchmark gain still supports after an overlap check.
ExampleHow to prepare
“You can look things up—just be comfortable with basic syntax and standard libraries so it doesn’t eat up your time.”
Anthropic CareersCareers page, "The interview process"
Last reviewed