Language Data Scientist II

The Bedrock AI Data Team in Amazon Web Services (AWS) is looking for a Language Data Scientist to collaborate in developing solutions for natural language data collections. This position is an opportunity to apply your expertise in a challenging but supportive environment.The mission of the Bedrock AI Data Team is to engineer the datasets critical to the success of AWS’s Bedrock services. From human evaluations to Responsible AI safeguards to Retrieval-Augmented Generation and beyond, these products make Generative AI enterprise-ready and safe for users, impacting millions of people every day. We are a group of language engineers, linguists, data scientists, data engineers, and program managers, and we partner closely with the science, engineering, and product teams. We are customer obsessed and committed to delivering results with the highest quality and integrity.As a Language Data Scientist, you will start by diving deep into a couple of critical projects for Bedrock services to drive these projects forward. You will collaborate with fellow language data scientists, program managers, as well as stakeholders in science, engineering, and product teams to understand the role data plays in developing models that meet customer needs. You will analyze, follow, and improve processes for collecting and annotating LLM inputs and outputs, assessing data quality, and automating where appropriate.You will then expand your scope by using the principles of data-centric AI to understand the role our data plays with regard to model performance specifically, as well as the larger ML pipeline. You will apply state-of-the-art Generative AI techniques to analyze how well our data represents human language and run experiments to gauge downstream interactions. You will work collaboratively with other language data scientists and scientists to design and implement principled strategies for data optimization.Key job responsibilitiesSource, validate, and deliver high-quality language artifacts and linguistic dataCollaborate with stakeholders to design data collection and development effortsOversee the progress and quality of several data collection and annotation projects at a timeAdvocate for strict adherence to data collection guidelines and quality thresholdsExtend existing data collection, annotation, and quality assurance efforts to support feature and language expansionInnovate on data collection methodologies, guidelines, quality metrics to support new requestsAutomate repetitive workflows and improve existing processesAbout the teamThe Bedrock AI Data Team at AWS is responsible for delivering high-quality annotated data and a variety of language artifacts to ensure the best performance of different AWS LLM services. These Generative AI services enable customers to readily add intelligence to their business operations and AI applications to drive positive outcomes.Minimum Requirements:2+ years of data scientist experience3+ years of data querying languages (e.g. SQL), scripting languages (e.g. Python) or statistical/mathematical software (e.g. R, SAS, Matlab, etc.) experiencePhD in Computational Linguistics, Linguistics with a computational component, or an equivalent field; alternatively, MA/MS with 3+ years of experienceExcellent knowledge of semantics, pragmatics, conversation analysis, and/or discourse analysisExperience designing and executing data collection projects, including guidelines, labelset and annotation workflow developmentExperience developing and evaluating data annotation and data quality metricsExperience designing and executing psychology/linguistic/cognitive science surveys or experiments with human participantsAdditional Preferred Qualifications:Ability to explain complex concepts and solutions in easy-to-understand termsExperience with synthetic dataset creationExperience with surveyingExperience working with a diverse array of languages or language varietiesPractical knowledge of version control systems such as GitHubAbility to unblock yourself in a fast-paced environmentExperience with driving cross-team consensus

#J-18808-Ljbffr