An Intelligent Web-Based Health Insurance Claim Processing System Using Generative AI, Retrieval Augmented Generation, and Optical Character Recognition
International Journal of Advanced Research in Science Communication and Technology · 19.05.2026 · 267
DOI: 10.48175/ijarsct-35538 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Advanced Research in Science Communication and Technology, online erschienen am 19.05.2026. Verfasst von Ritesh Yadav und 2 weiteren Autor:innen. Die Arbeit nennt 19 Literaturangaben.
Abstract (Originalsprache)
The development of responsive, production-ready web applications that integrate large language model inference pipelines presents a distinct set of software engineering challenges that extend beyond model selection and prompt design. This paper presents the design, implementation, and evaluation of a real-time AI-driven web application for automated health insurance claim adjudication, with specific emphasis on the frontend architecture, REST API design, asynchronous request handling, and prompt engineering methodology that collectively determine the quality of the user experience. The proposed system provides a professional dark-themed single-page web interface through which policyholders submit insurance claims by completing a structured form and optionally uploading medical bill documents. Submitted claims are processed through a Flask REST API backend that orchestrates optical character recognition for bill digitization, FAISS-based retrieval augmented generation for insurance policy lookup, and the Groq LLaMA 3.3 70B large language model for claim evaluation. A carefully engineered structured prompt template elicits consistently formatted five-section executive summary reports comprising a verdict, introduction, policy analysis, document verification, and conclusion. The application employs a three-stage pre-filter that resolves obvious rejection cases without invoking the language model, reducing average response time for rejected claims from seconds to milliseconds. The system is deployed as a Docker container on Hugging Face Spaces and evaluated through eight diverse clinical test scenarios covering covered conditions, excluded conditions, and financial boundary cases. Results demonstrate 100% verdict accuracy and median end-to-end response time of 3.2 seconds for claims requiring full language model evaluation. User interface design decisions including drag-and-drop file upload, animated loading states, color-coded verdict display, and quick-fill demonstration buttons are documented and evaluated for their contribution to system usability.
Zitieren
APA
Yadav, R., Swami, V., & Singh, R. (2026). An Intelligent Web-Based Health Insurance Claim Processing System Using Generative AI, Retrieval Augmented Generation, and Optical Character Recognition. International Journal of Advanced Research in Science Communication and Technology, 267. https://doi.org/10.48175/ijarsct-35538
Vancouver
Yadav R, Swami V, Singh R. An Intelligent Web-Based Health Insurance Claim Processing System Using Generative AI, Retrieval Augmented Generation, and Optical Character Recognition. International Journal of Advanced Research in Science Communication and Technology. 2026:267. doi:10.48175/ijarsct-35538
BibTeX
@article{Yadav2026,
author = {Yadav, Ritesh and Swami, Vandana and Singh, Rajendra},
title = {An Intelligent Web-Based Health Insurance Claim Processing System Using Generative AI, Retrieval Augmented Generation, and Optical Character Recognition},
journal = {International Journal of Advanced Research in Science Communication and Technology},
year = {2026},
pages = {267},
doi = {10.48175/ijarsct-35538}
}
EUVIMED