This Google exec is hiring for a new engineering team. She evaluates creativity in 2 ways during interviews.

• Google executive Yasmeen Ahmad said she is looking to hire engineers with fresh ideas.
• She said she has two methods of evaluating a candidate's creativity during a job interview.
• Ahmad said she looks for "early adopters" and those who can use AI to reimagine industries.

Google executive Yasmeen Ahmad is looking for something specific when hiring engineers — and it's not just technical know-how.

Ahmad told Business Insider that the typical software engineering interview used to focus on detailed coding tests and test suites. Now, as she hires for a forward-deployed engineering team, which will work with customers, she said she's prioritizing people with fresh ideas.

The strongest candidates are "able to think outside the box," Ahmad, director of Google Cloud's data cloud, said. "They're able to think outside the frame of how we would have normally described a problem."

The executive added that candidates who take a traditional approach to engineering aren't performing as well in her team's interviews. The ideal candidate nowadays, she said, can demonstrate creative problem-solving by using AI to reimagine traditional processes. She said she evaluates that type of thinking in two ways:

1. Constant experimentation

Ahmad said she looks for candidates who are constantly "tinkering" with new tools. That gives her an immediate signal that they're creative thinkers.

"When you're interviewing them, they're naturally immediately talking about, 'oh, last week I had tried AI in this context, and this is how it made me better at doing my job in this way,'" Ahmad said.

These candidates aren't trying new tools because their boss told them to or because it's the new cool thing to try, she said.

"They're the early adopters," Ahmad said.

Tech executives have told Business Insider that side projects are becoming increasingly common for candidates to demonstrate their aptitude in interviews. However, Ahmad said candidates don't need to have a GitHub repository of projects they've worked on in their spare time.

"It doesn't have to be pet side projects, because people are busy," Ahmad said, adding that workers can experiment on the job by trying out new ways to speed up their work.

2. Scenario testing

AI is being used more often throughout the interview process — in some cases, illicitly by job seekers, and in others, as a way for employers to test candidates' AI capabilities. As these tools reshape hiring, Ahmad said scenario-based testing has become central component to the interview process, giving hiring managers a better way to assess creativity.

Ahmad said she'll ask candidates how they would approach a scenario involving AI tools in an industry where they have no domain knowledge.

For example, if the example related to healthcare, a traditional candidate might say that they would take all the patients' unstructured PDFs, feed them into a single LLM prompt, and ask it to generate a summary for the doctor. That would be a "massive liability," Ahmad said, because in that scenario the candidate assumes AI can inherently understand the timeline of events or clinical context of an image by looking at it.

Ahmad said she's looking for a candidate who can "find solutions in a way that breaks the chains of how that workflow process has traditionally gone." So someone might suggest building the semantic context for the imaging data before the model sees it. Next, they would build a specific framework to ensure the agent is operating in the right time frame of data. Then, they would recommend designing a multi-step process that includes a continuous evaluation loop.

"We aren't just hiring people to write prompts," Ahmad said. "We are hiring people who can foresee how a model might silently fail in a high-stakes environment, and who know how to build the automated evaluation loops to catch it before it does."

She said asking these sorts of questions to vet creativity is especially useful as AI transforms the software engineering industry by automating core parts of the job.

"We're seeing the human role is evolving to more of an orchestrated role," Ahmad said. "So rather than having to write all of the detailed code, it's 'how do I actually express my intent to a multi-agent system now and have that multi-agent system execute on that intent?'"