Continuous Integration and Delivery of Machine Learning Models in Real-Time Financial Decision Systems

( Vol-12,Issue-3,May - June 2026 )

Author(s): Kaleshwar Aryasomayajula

Download Full Text PDF
Total View : 48
Downloads : 0
Page No: 394-400
DOI: 10.22161/ijaems.123.37

Keywords:

MLOps, CI/CD, machine learning models, financial decision systems, model risk management, credit scoring, data drift, real-time analytics, explainable AI, deployment governance

Abstract:

Real-time financial decision systems depend on machine learning models that change after release because input data, customer behavior, and portfolio risk move faster than traditional model governance cycles. This article examines continuous integration and delivery for models used in credit, fraud, eligibility, and limit decisions. The aim is to build an analytical framework for CI/CD in financial machine learning without presenting experimental claims. The study draws on ten recent sources on MLOps, data quality, concept drift, non-traditional credit data, and AI risk governance. Comparative source analysis, conceptual synthesis, classification, and analytical generalization guide the work. The results define three requirements: controlled pipeline promotion for code, data, features, and model artifacts; monitoring that separates data failures from model deterioration; and release governance that connects automation with audit evidence. Practitioners can use the framework to design model promotion, rollback, threshold review, and real-time trace logging for financial decision services under supervisory oversight.

Received: 20 May 2026; Received in revised form: 15 Jun 2026; Accepted: 19 Jun 2026; Available online: 23 Jun 2026

Cite This Article:

Share: