Open to opportunities

William Hart

@williamhart

Message

Senior AI Engineer and Data Scientist building production LLM/RAG systems that improve relevance and speed.

United States

Message

What I'm looking for

I’m looking to lead production GenAI/RAG and decisioning systems end-to-end—combining applied AI, experimentation, and backend engineering—to improve relevance, groundedness, and low-latency reliability in real products, with room to iterate and measure impact.

I’m a Senior AI Engineer and Data Scientist with 10+ years of experience building and operationalizing machine learning, experimentation, and decision systems across fintech, enterprise, and consumer platforms. I bring deep expertise in LLM workflows, retrieval-enhanced systems, causal inference, predictive modeling, risk modeling, ranking, and personalization—always with an emphasis on improving relevance, accuracy, latency, and reliability in production.

I combine applied AI, data science, and backend implementation to turn ambiguous business problems into measurable systems, products, and decisions. At Amazon, I shipped production LLM and RAG pipelines that improved response relevance by ~30%, increased transformer task accuracy by ~25%, and cut inference latency by up to 50% using vLLM, TensorRT-LLM, and Triton—backed by strong evaluation, A/B testing, and observability.

Experience

Work history, roles, and key accomplishments

Current

Senior Data Scientist - Applied AI

Current

Amazon

Sep 2022 - Present (3 years 8 months)

Built production AI systems for user assistance, workflow automation, and decision support using LLMs and retrieval pipelines. Improved response relevance by ~30%, increased transformer task accuracy by ~25%, and cut inference latency by up to 50% by optimizing serving and grounded RAG/Graph RAG workflows.

Retrieval Augmented Generation (RAG)Graph RAG VLLM Triton Prompt Engineering AWS Sagemaker

Data Scientist, AVP - AI/ML

Credit Suisse

Sep 2019 - Sep 2022 (3 years)

Owned ML services for personalization, ranking, and real-time decisioning in enterprise product workflows. Improved feature freshness by ~20%, reduced p95 latency by ~15%, cut user-facing errors by ~45%, and reduced API payload size by ~30% through low-latency scoring, caching/circuit-breakers, and GraphQL aggregation layers.

Personalized Ranking Java PostgreSQL Redis GraphQL Caching Timeout Controls React

Teaching Assistant - Data

Duke University

Jul 2019 - Sep 2019 (2 months)

Supported delivery of Duke’s Data Infrastructure course with implementation centered on SQL and Python for analytics-heavy coursework. Guided students on SQL/Python data pipelines and reinforced full-stack data workflows using React/TypeScript, REST APIs, GraphQL, Node.js, Docker, and AWS deployment practices.

SQL PostgreSQL Python fastAPI React TypeScript GraphQL REST APIs Docker

Co-Founder & Full-Stack Eng

Dough Pack Co.

May 2016 - Aug 2019 (3 years 3 months)

Built and launched a peer-to-peer student lending platform in Taiwan, owning product design and full-stack implementation from concept to release. Developed backend services with Node.js/Express and SQL databases, and applied Python/Pandas/SQL pipelines with risk scoring and matching logic to improve lending decisions.

Express REST APIs PostgreSQL MySQL Python Pandas SQL Risk Scoring API Design NodeJS

Education

Degrees, certifications, and relevant coursework

Duke University

Master of Science in Data Science, Data Science

2017 - 2019

Completed a Master’s degree in Data Science at Duke University, building expertise in statistical modeling, machine learning, data analysis, and quantitative problem-solving through rigorous training and applied projects.

Duke University

Master in Interdisciplinary Data Science (MIDS), Interdisciplinary Data Science

2016 - 2017

Pursued the Duke Master in Interdisciplinary Data Science (MIDS) to transition into data science and AI, strengthening interdisciplinary depth across quantitative analysis, modeling, and implementation-oriented technical work.