Research Scientist Amazon
Berkeley, CA

Job Description

Next Generation of Amazon Search

Amazon is the 4th most popular site in the US (source:

http://www.alexa.com/topsites/countries/US). Our product search engine,

one of the most heavily used services in the world, indexes billions of

products and serves hundreds of millions of customers world-wide. We are

working on a new initiative to transform our search engine into a

shopping engine that assists customers with their shopping

missions. We're looking at every aspect of search, from query

understanding to front-end UX, ranking and relevance, indexing and

tiering and asking how we can make big, step improvements by applying

advanced Machine Learning (ML) and Deep Learning (DL) techniques. This is a rare opportunity to develop cutting edge ML solutions and apply

them to a search problem of this magnitude. Some exciting questions that

we expect to answer over the next few years include:

* Can we deeply understand customer intent and personalize their search

experience even when they type broad queries such as "dress" or

"espresso machine"?

* Can we reduce the cost of serving customer queries on Amazon by two

orders of magnitude using ML to predict n-grams and tuples that many

queries decompose into, apply expensive ranking functions offline to

identify the most relevant products that match these terms, and index

these for efficient online retrieval? We expect this to lead to exciting

research at the intersection of systems and ML.

* Can we deeply understand the catalog to surface products that offer

the most value to a customer? The challenge here is that the definition

of value is subjective and personal, and therefore requires a deeper

understanding of the customers intent as well as preferences.

* Can we increase the experimental velocity of Customer Experience (CX)

experiments by two orders of magnitude? Achieving this will enable us to

rapidly try various CX treatments, and contextualize the CX based on

factors such as customer intent and device.

* Can we use deep learning to transfer behavioral signals from

frequently purchased products in the head to products in the tail where

behavioral signals are sparse? The challenge here is the scale, and the

fact that the head and torso contain only a small fraction of products

while the tail contains an overwhelmingly large fraction of the products

in the catalog.

We are looking to hire ML Applied Scientists at all levels, with experience in Search, Personalization, NLP, Systems, ML, DL and UI Design. Internship opportunities are also available throughout the year and we are flexible about duration and start dates. You will be working alongside world-class researchers and engineers to build next generation search systems and will be able to deploy your ML models into production. Our team is proud of its collaborative and open research environment, where long term thinking and risk taking are highly rewarded. We value academic collaborations and encourage our scientists and engineers to participate and publish in top conferences such as NIPS, ICML, KDD, SIGIR and WWW.

Positions are available in the new Amazon office in Berkeley, CA.