Google DeepMind is hiring for Polaris.
Total compensation runs from $550,000 to $1.1 million a year, and you don’t need a degree to apply. The base salary alone sits between $164,000 and $252,000. Google first called for Polaris applicants on September 25, describing it as a new DeepMind team designing evaluation frameworks for AI coding agents.
“All we care about is you being smart and able to solve hard problems,” he wrote, adding that people should ignore the pay mentioned in the job ad. Google DeepMind is hiring engineers for Polaris, a new team that builds tests to measure how well Gemini handles real software engineering work. The pay stands out. Tamay Besiroglu, an Epoch AI co-founder who now works on data at Google DeepMind, kept the brief simple in a post on X. When asked whether the listing was wrong, he clarified that its numbers were accurate but left out equity. The role, listed as Research Engineer and based in Mountain View, California, centres on designing tasks. Each one is a detailed exam for Gemini, made up of a prompt telling the model what to do and a grader or rubric that scores its attempt. Engineers will also start new projects to improve Gemini’s training data and evaluations. On paper, the minimum bar asks for software development experience and hands-on use of AI coding tools and developer agents. A track record at contests like the IMO, Putnam or IOI, or experience designing benchmarks and evaluation rubrics, counts in your favour.