University of North Texas
Endeavor: A Regulated AI Workbench for Board Order Coding and Review
Abstract
dc:descriptionHarris, Reyden. Endeavor: A Regulated AI Workbench for Board Order Coding and Review. Master of Science (Data Engineering), May 2026, 86 pp. 4 appendices, 13 Figures, 15 references. This thesis presents the design, implementation, and evaluation of an AI-assisted document analysis system to support Board Order coding within state medical board regulatory workflows. Board Orders document disciplinary actions taken against medical practitioners and must be interpreted, coded, and reported in accordance with national standards to ensure public safety and interstate regulatory coordination. The existing Board Order coding process relies heavily on manual review of lengthy legal documents, introducing delays, variability, and scalability limitations. While large language models and document intelligence systems offer promising capabilities for text extraction and summarization, their unconstrained application is ill-suited to compliance-sensitive environments that require determinism, auditability, and human oversight. To address these challenges, this thesis introduces a production-grade, human-in-the-loop workbench that integrates document ingestion, optical character recognition, structured large language model-based extraction, and governed configuration management. The system produces schema-constrained JSON outputs aligned with national reporting requirements and presents results through an interactive web interface designed for regulatory review. An empirical evaluation compares legacy OCR tooling with managed document intelligence services, demonstrating substantial improvements in processing latency, extraction fidelity, and consistency. Similarity-based metrics and qualitative review by domain experts from state medical boards are used to assess system suitability in practice. The results demonstrate that AI-assisted document analysis can meaningfully support regulatory workflows when deployed within a constrained, auditable architecture. The system reduces administrative burden, improves extraction reliability, and preserves human authority in high-stakes decision-making. These findings establish the feasibility of integrating generative AI into medical board regulatory processes while maintaining compliance, accountability, and public trust.
Degree
thesis:*- Grantor dc:publisher
- University of North Texas
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Harris, Reyden, 1993-
- Contributors dc:contributor
-
- Murali, Beddhu
- Fu, Song
- Struen, Cynthia
Subjects
dc:subject × 3Rights
dc:rights- Statement dc:rights
-
- Public
- Harris, Reyden, 1993-
- Copyright
- Copyright is held by the author, unless otherwise noted. All rights Reserved.
- Language dc:language
- English
Identifiers
dc:identifier.*- Identifier
-
local-cont-no: submission_4842
ark: ark:/67531/metadc2588693