Vanessa Murdock is a classical pianist turned Amazon applied scientist

Vanessa Murdock is a manager of Applied Science on the Amazon Alexa Shopping team. Vanessa is a trained classical pianist turned information retrieval researcher — by no means your typical career journey. In this interview, Murdock talks about how her training as a classical pianist helps her be a better scientist, why she joined Amazon, and how her work at Amazon affects the lives of millions of customers.

Tell us a little about your background.

When I was younger, we lived with my grandparents for a few years. My grandfather was a labor lawyer. People often paid him in forms other than cash, and one of his clients paid him with a Steinway piano. I started by sounding out melodies on the piano, playing by ear. When I was four, my grandmother heard me piecing together a Mozart Symphony with both hands and a harmony, so she started to teach me. I didn’t take the lessons very seriously.

When I was 12, I broke my leg. I was bored because I had to stay inside a lot. As a result, I started practicing three or four hours a day. The difference between how well you play when you practice 20 minutes a day and when you practice four hours a day is vast. I started winning piano competitions, including one that sent me to Europe to give a concert at a music festival. Eventually, I received a scholarship at Texas Christian University (TCU), which hosts the Van Cliburn International Piano Competition. I went there to study with a Van Cliburn winner, Steven DeGroote – the Van Cliburn piano competition is held every four years; winners and runners-up receive cash prizes, in addition to the opportunity to perform at world-famous venues

What does it take to be a really good pianist?

You have to be very analytical and self-critical if you want to be a good pianist. You have to learn to hear how you sound as if you were sitting in the audience, and to be thoughtful about all the little choices you make. No detail is too small. Being analytical and self-critical have helped me a lot in computer science, at Amazon and in life in general.

How did you get into information retrieval?

When I started my career as a pianist, I took other work (I was a bookstore employee, I did housekeeping and food service, I worked in a dry cleaner) to supplement my income. As I became more established in the city I was living in, I was able to make a living solely from music jobs. I played in musical theaters, at weddings and parties, in churches, I taught privately and at a private school where I was also the staff accompanist, I performed as a soloist with orchestras and I had a trio that played concerts.

Although I was successful as a pianist, I was working 50 hours a week or more, and I was still struggling financially. When my son was born, it became clear that he would have fewer opportunities than I had, because I would not be able to give him a middle-class upbringing with extras like sports and music lessons. I was also a little burned out on teaching and accompanying. The part of music I loved was performing classical music, but I did not derive enough income from performing to do only that. I decided that I had to change paths.

I looked at a number of fields like journalism, political science and labor law. However, although they were interesting and would have been engaging, they also had long hours and low salaries. Then one day I was chatting with a friend on AOL messenger, and I started thinking about the magic of instant messaging: you can type a message and in an instant another person can read and respond, regardless of where they are in the world. I decided that I wanted to learn how computers work. My plan was to take a day job as a programmer, figuring it would provide a steady income with health insurance. It would only be 40 hours a week, which would leave me more time to focus on my performance career.

I enrolled at Colorado State University. To my great surprise, computer science was extremely fun, and much easier than piano. In the summer before my senior year, I took an internship at AT&T Research in New Jersey, working on machine translation with Srinivas Bangalore. The project was to mine the Web for parallel texts to train a machine translation system. A week into the internship I had an epiphany that computation was a tremendously powerful tool to understand fundamental questions about humanity, and I was hooked.

It was that internship, and Dr. Bangalore’s mentoring that showed me that instead of taking a “day job” testing printer drivers, I could do something really enriching. I was very fortunate that Dr. Bangalore encouraged open-ended exploration of the research questions. I had lofty goals at the time because I was inspired and idealistic, but I still find the big open questions about how people understand information to be the most compelling.

I decided that I wanted to do research, so I pursued a PhD. AT&T gave me a grant which included ongoing mentoring from Charles Thompson, who was on the board of the AT&T Fellowship program. Dr. Thompson helped me to understand that AT&T was supporting me because they saw in me a world class researcher. The combination of Dr. Bangalore’s big thinking and Dr. Thompson’s steady insistence that I could do significant science really changed the game for me. The lessons from the two of them infuse all of my work and all of my mentorship of new researchers.

Why did you join Amazon?

I am really excited about cloud-based voice services because voice will ultimately be a natural way for people to interact with their devices. Voice interfaces give us another picture of how people communicate. I like Amazon’s obsession for looking at problems from the customer perspective, and the potential to use science to directly improve the lives of millions of people.

The projects that I find most inspiring are the ones that allow me to understand customers better. My team is working on understanding what products are potentially embarrassing, and finding ways to be sensitive to these issues when providing experiences for our customers. For example, I don’t mind if people know I dye my hair because my hair is blue but another customer might be embarrassed if Alexa recommends dye with “full coverage for grey hair” in response to their shopping request.

I also love projects where we can help customers find what they are looking for or save them time. For example, people often reformulate their discovery query when they are not satisfied with their results. They might start by querying for “latte,” before reformulating their query to “espresso machines” to get more relevant results. My team’s research allows us to build experiences that help our customers find what they are looking for faster.

What’s different about working at Amazon?

One thing that’s really different at Amazon is how we discuss ideas and plans as a document that everyone reads through together. This seemed like overkill the first time I saw it, but a couple weeks in, I realized that a six-page narrative is a great equalizer. When ideas are presented verbally, they can be less convincing if the presenters are not skilled, or unduly credible if the presenters are charismatic and able to charm the audience into supporting a weak idea. Further, the audience may think they agree with a proposal, but actually misunderstand it, leading to serious friction down the line. Having the information presented as a document resolves much of this because the document is concrete and it can be edited to be clearer, and referred back to when there are questions later. If all the stakeholders agree on the substance of the document, it becomes their contract. It is the most effective way I have seen to come to an understanding as a group and make a rigorous group decision.

Amazon is optimized for shipping innovations quickly – the amount of time to go from first idea to customer-facing product is much shorter than at other places I have worked. People show genuine excitement and energy for what they are doing, and what they could do in the future. Everyone is completely focused on making a meaningful difference for customers. As a result, many good decisions are baked into our mechanisms, rather than being the result of an afterthought.

As scientists, our best ideas come from a deep understanding of a problem. You can have a certain depth of understanding by reading papers and running simulations, but it does not compare to the depth of understanding you gain from making scientific advances on real systems that are useful and relevant to people. The change from music to computer science was a huge change, but being at the front of a technological revolution is exciting and I am honored to play a part.