Editor’s note: Belinda Zeng joined Amazon in 2017 as the global head of data science and has participated in hundreds of interviews for science roles across the company. Here she shares her thoughts on what it takes to succeed as a scientist at Amazon.
I have had the pleasure of working at Amazon as a science leader for the past four-plus years. Two years ago I became what is known in Amazon as a Bar Raiser. Bar Raisers are experienced interviewers who help to raise the Amazon recruiting standard. I lead a science and engineering team called M5 — the five Ms stand for multi-lingual, multi-locale, multi-modal, multi-task, multi-entity — a large-scale AI program focused on transforming how deep learning models are built and deployed at Amazon. My team innovates to help bring Amazon services beyond the current state of the art, achieve step function improvement, and unlock many new downstream applications in search, advertising, and catalog, to name just a few.
Looking back on my journey at Amazon, and drawing on my experience as a Bar Raiser, I’d like to share some information and advice with those who are interested in exploring opportunities with Amazon.
What does the hiring team look for?
I still remember the day when I submitted my application to Amazon, wondering what the hiring team was seeking. Four years later, I know the answer to that question.
First and foremost are the functional competencies, including science breadth, depth, experience in developing science applications, and scripting language coding skills. There are a number of science roles within Amazon and because the core responsibilities for those roles are distinct, the required technical skills differ.
Data scientists, for example, are considered as generalists who investigate the feasibility of applying scientific principles to business problems. They are normally assessed for data skills, math/stats knowledge and, most likely, analytical mindset, and business acumen.
Research and applied scientists are expected to have deep expertise in one of the data-driven science disciplines and to apply scientific principles to support significant invention. The hiring team typically delves into one or two scientific areas such as machine learning, speech recognition, operations research, and robotics.
Development of software code is a core skill expected from applied scientists as they are deeply involved in bringing their algorithms to production. Economists are vetted for their experience developing offline code for applied econometric applications. The second area we assess is how well applicants can apply the Amazon Leadership Principles. In the more than 200 loops (Amazon’s name for our interview process) in which I have participated, three Leadership Principles stand out for scientists:
- Learn and Be Curious: In my interview conversations, I look for data points that show the candidate proactively seeking opportunities to learn new skills and improve themselves versus staying with familiar situations or avoiding new experiences.
- Dive Deep: I look for those who investigate and get details to solve a problem, even when faced with challenges, as opposed to having only a surface-level understanding of projects;
- Invent and Simplify: I look for those who generate new ideas or simplify a solution for long-term wins versus creating a cumbersome process to solve a short-term problem.
For senior level roles, a writing exercise is normally required as well. Amazon uses written documents to communicate ideas and influence others. We look for candidates who are able to articulate a process, product or point of view in a clear, crisp, and logical manner.
During the interview debrief, we often debate whether a candidate “raises the bar”. A bar-raising candidate is a candidate who is better qualified than 50% of existing employees at the same level. For entry level roles, it means the ability to fulfill a task with supervision. For experienced hires, it means to deliver with autonomy and minimum supervision.
How does Amazon support its scientists?
For scientists hired by Amazon, there are many types of career support available from both your team and the company.
Learning: Amazon seeks candidates who are passionate lifetime learners, and provides numerous opportunities to support that instinct. That can come in the form of online and classroom courses, team wiki and learning portals, as well as access to experts and mentorship. For example, 200 Amazon scientists were randomly selected to participate in a Coursera beta program to take free online courses for six months. The scientists were able to stay current in their science specialty and increase their skills and knowledge to apply on their job.
In addition, there is a special program called the Day 1 Science Mentorship Program. That program pairs up new-hire scientists with experienced Amazon science leaders to ease the transition into Amazon’s business culture.
Community connection: An expansive community is critical to a scientist’s development. At Amazon, there are hundreds of science-focused meetings, reading clubs, invited talk series, and workshops happening on a regular basis. These mechanisms not only offer the opportunity to connect with people who have similar research interests, but also provide a forum to showcase innovative work.
The company also holds multiple annual science conferences for Amazonians interested in innovative science. One is the annual Amazon Machine Learning Conference, a four-day event that covers most major areas in machine learning and attracts thousands of attendees and submissions. Collectively we continually raise the scientific bar at Amazon.
Growth: At Amazon, we all grow with the company. There are ample opportunities to stretch yourself, by expanding your scope and growing your skill set. I have helped scientists on my team transition into different science roles; relocate internationally for a stretching assignment; and watched some go from individual contributors to tech leads and eventually managerial positions.
How do you build a successful career at Amazon?
Here are some insights from my personal experience:
Trust is a multiplier. There are multiple meanings inside this single word: transparency, integrity, capability, and many more. For scientist roles, trust naturally expands with competency — stay fresh, relevant and capable — and contribution, which means producing high quality, timely results. I have worked with many great scientists and observed how they build trust through capability and results, which in turn brought greater influence. A common pitfall is sometimes we tend “assume” trust by overestimating our capabilities. Consistently asking for feedback, then listening to and acting on that feedback will help close that gap and build trust.
Work backwards from a problem. New scientist hires, especially those who recently moved from a foundational research role, sometimes find it hard to transition into the Amazon working backwards culture. The goals in foundational research are to generate knowledge or understanding regarding a particular phenomenon, without much focus on real-world impact. However, for applied research at Amazon, the main criterion of success lies in how well findings can be used to have a positive impact on customers. A well-balanced focus between curiosity- and solution-driven research is key to ensure effective execution.
Be a well-rounded scientist. Being a scientist means more than running experiments. Scientists are expected to understand the business problem, decompose a complex issue into components that are addressable by science, and communicate science effectively. Success is the journey, not the destination. If you are interested in joining Amazon’s customer-obsessed journey, please visit the Amazon Science careers page. It is always Day 1 at Amazon.
