Applied Scientist II, Sheriff Team- Payroll tech

Payroll Tech's Sheriff team develops and maintains ML and Generative AI applications that support Payroll Operations and Amazon employees at scale. Our portfolio includes Pay-Input Anomaly Detection, which improves the pay experience by identifying pay input irregularities such as leaves and insurance discrepancies; Percept, which improves ticket resolution by providing intelligent ticket prioritization via sentiment scoring, ticket summarization, defect classification, and categorization; Penny, a Virtual Assistant that enables payroll operations teams to efficiently retrieve information from multiple sources including policies, Percept data, vendor data, and HR data via Xylem through a single browser interface; Pay Ticket Genie, in process of being integrated with Amazon AZA(A to Z Assistant); Niyam, our rule engine; and Policy as Code Extraction (PoCo), a critical component of SPACE (Single Payroll Autonomous and Computation Engine) Amazon's in-house payroll system built to eliminate third-party vendor dependency for payroll processing. PoCo ensures data accuracy by validating that pay instructions are correct and performing calculations when required. It consists of two components: policy-based rule creation, where business owners select a pay code and provide policy links to generate rules for specific business processes, and rule evaluation, where upstream services send real-time validation or calculation requests and receive results along with rationale for any failures. Sheriff team owns policy-based rule creation and powering the rule evaluation system with rules generated.
As an Applied Scientist on the Sheriff team, you will own and advance the ML and GenAI capabilities that power these systems: driving model accuracy, scientific innovation, and global scale across the payroll ecosystem.
Key Job Responsibilities
As an Applied Scientist on the Sheriff team, you will operate across three core dimensions: Invent, Implement, and Influence.
Invent
You bring deep domain knowledge and fluency with state-of-the-art scientific approaches as well as emerging technologies from the research community. You practice customer-obsessed science : working backwards from the needs of Amazon employees and payroll operations teams to extend or invent new ML approaches, even when no textbook solution exists. You design novel ML and LLM-based methodologies for anomaly detection, sentiment analysis, ticket classification, prescriptive analysis, intelligent virtual assistance, and automated policy extraction. You identify and define the research agenda for expanding Percept's capabilities including prescriptive analysis feature; lead the scientific strategy for the Penny-AZA integration enabling accurate and low-latency responses to Amazon employee payroll queries, and drive the ML strategy for Policy as Code extraction(PoCo), developing models that extract, interpret, and codify payroll policies into structured, executable rules that power real-time pay instruction validation and calculation within SPACE. You author or co-author articles for internal or external peer-reviewed venues that validate the novelty of your work, when appropriate and not precluded by business considerations.
Implement
The ML components you develop are directly integrated into production systems or directly support large-scale applications serving Amazon's global payroll operations. You make appropriate tradeoffs between model accuracy and latency, innovation and stability, and immediate versus long-term solutions; favoring reuse and established frameworks where appropriate. You make progress semi-autonomously with only occasional guidance, implement at the correct level of complexity the first time, and evaluate emerging technologies including Large Language Models (LLMs) and GenAI frameworks largely on your own. You ensure your models and pipelines integrate robustly with data sources including USC (Unified Central Service), Xylem, SIM-Ticketing, Pay Code Governance system, and PoCo's rule evaluation engine.
Influence
You contribute to tactical and strategic planning for the Sheriff team, including goals, priorities, and roadmaps for ML and GenAI capabilities. You lead the scientific strategy for the global expansion of Percept, driving both feature growth and country-level launches own the complex AI track for Penny-AZA integration collaborating across partner teams including Reflect and GREF to ensure seamless data integration and robust ML pipeline workflows, and drive the scientific roadmap for PoCo's expansion to 100K US employees, ensuring the ML models powering policy extraction, rule generation and testing scale reliably to meet this growth. You mentor scientists and engineers on the team and across teams, championing best practices for the AI-Driven Development Life Cycle (AIDLC) to rapidly increase developer productivity and delivery velocity. You provide peer feedback on research procedures and results within and across teams, help recruit and develop bar-raising talent through interview drives, and grow the team's organizational knowledge of Sheriff team ML solutions. You are visible in the broader internal and external scientific communities as a subject matter expert and regularly serve as a Program Committee (PC) member at peer-reviewed conferences or review articles for journal publications.
- 3+ years of building models for business application experience
- PhD, or Master's degree and 4+ years of CS, CE, ML or related field experience
- Experience in patents or publications at top-tier peer-reviewed conferences or journals
- Experience programming in Java, C++, Python or related language
- Experience in any of the following areas: algorithms and data structures, parsing, numerical optimization, data mining, parallel and distributed computing, high-performance computing
- Experience using Unix/Linux
- Experience in professional software development
- Experience communicating research findings and analysis in both written and spoken channels
- Experience with Machine Learning and Large Language Model fundamentals, including architecture, training/inference lifecycles, and optimization of model execution, or experience leading and influencing your team or organization
- Experience in machine learning, data mining, information retrieval, statistics or natural language processing, or experience in developing and deploying LLMs in production on GPUs, Neuron, TPU or other AI acceleration hardware
- Experience with training and deploying machine learning systems to solve large-scale optimizations, or experience with data infrastructures: relational analytic DBMS, Elastic-Search, and Big Data EMR/EC2/Glue/Lambda
- Experience using strong customer service, communication, and interpersonal skills
Our inclusive culture empowers Amazonians to deliver the best results for our customers. If you have a disability and need a workplace accommodation or adjustment during the application and hiring process, including support for the interview or onboarding process, please visit https://amazon.jobs/content/en/how-we-hire/accommodations for more information. If the country/region you’re applying in isn’t listed, please contact your Recruiting Partner.