Microsoft Senior Applied Scientist Web Data 

China, Beijing, Beijing 

623428237

At the Web Data team we're looking for a passionate Applied Scientist to help us on our mission of employing deep learning to understand all the data on the web; the largest store of information in human history. With this understanding we power end-user experiences across a variety of areas, including:
• Search relevance
• Recommendation system

As a team, we play a central role in document understanding and aggregation of data across documents leveraging the web graph. Work in our team is unique in that you'll be able to work with trillions of documents, which pushes the boundaries of our infrastructure and computational resources. Moreover, it provides interesting challenges that haven't been encountered before, thereby combining research and product applications.
Examples of our team's work include:
• Revolutionize the field of information retrieval through bridging the gap between end to end dense retrieval and the traditional cascade approach of sparse retrieval with BERT reranking.
• Establish new benchmarks for document understanding tasks
We are looking for a passionate Applied Scientist with demonstrable skills in information retrieval, deep learning, natural language processing, large-scale distributed computing, and LLM fine tuning

• Minimum: Bachelors; Preferred: advanced degree and/or industry experience
• Experience applying deep learning techniques, LLM fine tuning
• 7+ years of experience coding in Python, C++, C#, C or Java
• Passionate and self-motivated
• Ability to effectively collaborate and ship production features in a multi-project, fast-paced team environment
• Good communication skills, both verbal and written
• Focus on customer impact during design and development
• Ability and motivation to self-teach while entering new domains and managing through ambiguity

• Research and development of deep learning models for Microsoft AI scenarios, including language understanding, multi-modal representation learning, etc.
• Defining problems and establishing metrics
• Gathering training data at scale
• Exploring model design and architecture
• Exploring learning objectives and tasks
• Optimizing solutions and shipping to production
• DL infrastructure and platform support