Applied Scientist

Job ID: 1234167 | Services LLC


The Amazon Fashion catalog spans millions of items and variations ranging from high fashion to everyday wear. With a large, varied, and growing selection, we are looking for innovative ways to build a catalog that can drive product discovery, personalization, and a world-class customer experience.
Downstream services such as recommendation engines, browse, and search experiences are dependent upon high quality and consistent catalog attributes. To optimize these inputs, we will need to manage massive quantities of data and find novel ways to apply statistics and machine learning to complex business problems.
The Amazon Fashion team is looking for an experienced Applied Scientist to apply machine learning solutions to these business problems and connect customers to the products they love. You will be at the intersection of statistics, machine learning, data visualization, and business/product management. You will work closely with a multidisciplinary team to design solutions that impact millions of customers, sometimes from scratch, and apply models and algorithms in real-time systems at a very large scale.
The ideal candidate is a specialist in the Computer Vision (CV) domain and has a strong understanding of supervised, unsupervised and weak learning systems to quickly scale model training and inference. The candidate should be familiar with state-of-the-art techniques such as ResNet. Apart from the specialized domain, candidate should have sufficient understanding of deep learning techniques which includes applications for NLP models. They should also be a general athlete in data with broad exposure to experimental design, modeling, working with large-scale unstructured data to uncover key insights. They should be able to ask the right questions, prioritize, and communicate/align across both technical and non-technical stakeholders. This candidate exemplifies bias for action, the desire to continuously invent and simplify, and most importantly, a genuine curiosity and passion for learning.

Roles and Responsibilities:
· Explore unstructured data and determine catalog improvement opportunities
· Investigate the feasibility of ML or statistical techniques to business problems and products
· Rapidly prototype with off-the-shelf machine learning models to establish baseline models
· Collaborate with other scientists across teams to develop state-of-the-art performing algorithms and present technical findings to the larger science community through meetings/conferences/publications
· Drive best practices, helping to set high scientific and engineering standards on the team.
· Work with engineers to build requirements and productionalize models and pipelines
· Communicate requirements in a way that is consumable by engineers and business stakeholders


· PhD or equivalent Master's Degree plus 3+ years of experience in Computer Science, Machine Learning or related field
· Extensive knowledge of state-of-the-art CV techniques such as ResNet
· Extensive knowledge of unsupervised learning (such as product clustering using text embeddings)
· 2+ years of experience of building machine learning models for business applications
· Experience programming in Java, C++, Python or related language
· Knowledge of software engineering best practices including coding standards, code reviews, source control management, build processes, testing, and operations
· Strong verbal/written communication and data presentation skills, including an ability to effectively communicate with both business and technical teams
· Experience conducting large scale data analysis to support business decision making
· Proficiency in manipulating unstructured datasets using Python/SQL


· Extensive knowledge and practical experience in several of the following areas: machine learning, statistics, deep learning, Computer Vision, Natural Language Processing;
· Significant peer reviewed scientific contributions in premier journals and conferences
· Strong fundamentals in problem solving, algorithm design and complexity analysis