# American Express Data Scientist Jobs and Interview Questions

- AmEx plans to hire 20 data scientists in the next 6-12 months.
- The new hires will work on a machine learning setup that is one of the largest in the financial industry.
- Interviewers want candidates who demonstrate skill and learning agility in their problem solving.

*If I have two children and one child is a girl, what is the probability that both children are girls?*

For most people, this question is probably a stumbling block. But if that sounds obvious, you might have what it takes to fill one of the 20 data scientist positions American Express is hiring for over the next six to 12 months.

“Solving this problem is no different from how we build a machine learning model and build the mathematical model to solve a real problem,” said Di Xu, vice president of AI Labs and Governance. of AI at AmEx.

AmEx’s push for data scientists follows a rise in industry-wide demand for similar roles. According to the Bureau of Labor Statistics, employment in data science and related occupations is projected to grow 31.4% from 2020 to 2030. And AmEx’s own transition over the past decade from statistical learning models traditional to machine learning leads with broader use of business processes.

AmEx began investing in AI in 2010, according to Anjali Dewan, vice president of consumer marketing and business personalization decision science, in hopes of seeing if the model would better help address the roughly eight billion business transactions.

It did, and as of 2014, 100% of the company’s credit and fraud responders moved to AI.

When interviewing for data scientist roles, Xu said interviewers try to understand a candidate’s skills, learning agility and adaptability, in order to make AI smarter and more efficient.

The interviews are divided into two sections to test technical skills and behavioral skills. Xu said the questions are more open-ended so interviewers can observe how the candidate goes through the problem-solving process.

“The final answer, whether right or wrong, matters less,” Xu said. “We care more about the process by which the candidate solves the problem.”

Some questions don’t have definitive answers, but those that do can be answered in different ways, Xu said. “We can design additional questions along the way to probe how the candidate will define the problem and answer the question.”

For example, candidates usually start by answering a simpler question, such as the one mentioned above:

*If I have two children and one child is a girl, what is the probability that both children are girls?*

Some candidates will approach this probability-based question by using mathematical theory to find a formula; or they can use computer code that can solve the problem. But Xu said the process was more complex.

“There is a way for people to define this problem in a very concise and simple way,” Xu said. “If they can do that, it’s no different than creating a little mathematical model around the problem.”

The next step in this question adds another dimension: *If I have two children, one of them is a girl born on Tuesday. What is the probability that both children are girls?*

If the contestants can properly frame this more complex issue, Xu said it’s helpful to understand how they approach the issues.

Other prompts in the interview process are more hypothetical, like this: *Design some kind of data product for a restaurant – we need to know the prices of entrees on the menu of a given restaurant. How do we use transactional data to infer this?*

Transactional data is among AmEx’s most important data, Xu said, because it can display a lot of information and allows AmEx to continue innovating and improving the customer experience.

Xu said he didn’t know the perfect answers to the hypothetical questions himself — he wanted to see what the contestants came up with instead.

“They can give us ideas, and maybe the perfect answer can come out of that process,” he said.

In the behavioral part of the interview, Dewan said interviewers want to see if candidates have leadership skills. Candidates should be effective collaborators, working in different groups to better understand issues.

According to Dewan and Xu, what matters most is the speed at which a candidate learns and their willingness to improve.

“Competence is only one aspect, but learning agility is the other aspect,” Xu said. “It’s very important for our future success.”