The accelerating digitalization of retail enterprises has placed unprecedented operational strain on legacy infrastructure that was never designed to support real-time analytics, continuous deployment, or machine learning–driven decision systems. As retail organizations increasingly integrate machine learning models into mission-critical workflows such as demand forecasting, pricing optimization, fraud detection, and inventory management, the reliability of both software systems and embedded models has emerged as a strategic concern rather than a purely technical one. Site Reliability Engineering (SRE), originally conceptualized within large-scale cloud-native organizations, offers a principled approach to managing reliability through explicit service-level objectives, automation, and error budgeting. However, the translation of SRE principles into legacy retail environments—characterized by monolithic architectures, heterogeneous data pipelines, and organizational inertia—remains insufficiently theorized. This research addresses that gap by developing a comprehensive socio-technical framework that integrates SRE practices with modern machine learning observability and MLOps methodologies in the specific context of legacy retail infrastructure.

Drawing on an extensive critical synthesis of contemporary scholarship on SRE implementation in retail systems (Dasari, 2025), machine learning drift detection, model monitoring, observability, industrial reliability engineering, and ethical AI governance, this article advances a holistic conceptual model for resilient retail operations. Rather than treating infrastructure reliability and model performance as separate domains, the study conceptualizes them as co-evolving layers within a single operational ecosystem. The methodology is grounded in qualitative comparative analysis of documented industry practices, theoretical extrapolation from reliability engineering literature, and interpretive analysis of production ML monitoring frameworks. The results demonstrate that reliability failures in retail systems are rarely attributable to isolated technical faults; instead, they emerge from systemic misalignments between organizational incentives, data quality regimes, observability maturity, and operational governance structures.

The discussion critically evaluates competing scholarly perspectives on automation, human oversight, and ethical accountability in high-reliability digital systems, arguing that SRE-informed MLOps provides a viable pathway for balancing operational resilience with responsible AI deployment. The article concludes by articulating implications for practitioners, researchers, and policymakers, emphasizing the necessity of reimagining legacy retail infrastructure as adaptive socio-technical systems rather than static technological artifacts. By embedding machine learning observability within an SRE-oriented reliability culture, retail organizations can transition from reactive incident management to proactive, ethically grounded operational excellence.

Site Reliability Engineering, Legacy Retail Infrastructure, Machine Learning Observability, MLOps

Encord. (2024). A guide to machine learning model observability. Encord Blog. https://encord.com/blog/model-observability-techniques/

Dasari, H. (2025). Implementing site reliability engineering (SRE) in legacy retail infrastructure. The American Journal of Engineering and Technology, 7(07), 167–179. https://doi.org/10.37547/tajet/Volume07Issue07-16

Lewis, G. A., et al. (2022). Augur: A step towards realistic drift detection in production ML systems. ACM. https://insights.sei.cmu.edu/documents/614/2022_019_001_877199.pdf

Google Cloud Platform. (2021). Understanding machine types. https://cloud.google.com/compute/docs/machine-types

Maverick, V. (2019). Log management best practices: A comprehensive guide. Loggly Blog. https://www.loggly.com/blog/log- management-best-practices-acomprehensive-guide/

Payette, M., & Payette, M. (2023). Machine learning applications for reliability engineering: A review. Sustainability, 15(7). https://www.mdpi.com/2071-1050/15/7/6270

UTS Data Science Institute. (2020). Ethics of AI: From principles to practice. https://www.uts.edu.au/globalassets/sites/default/files/2021-02/executive-summary-of-ethics-of-ai- fromprinciples-to-practice.pdf

Shopify Engineering. (2019). Observability at Shopify. https://engineering.shopify.com/blogs/engineering/observability-at-shopify

Huang, B., et al. (2020). Modern machine learning tools for monitoring and control of industrial processes: A survey. ResearchGate. https://www.researchgate.net/publicat ion/341763531_Modern_Machine_Learning_Tools_for_Monitoring_and_Control_of_Industrial_Processes_A_Survey

Singla, A. (2023). Machine learning operations (MLOps): Challenges and strategies. Journal of Knowledge Learning and Science Technology, 2(3). https://www.researchgate.net/publication/377547044_Machine_Learning_Operations_MLOps_Challenges_and_Strategies

Evidently AI. (2025). Model monitoring for ML in production: A comprehensive guide. https://www.evidentlyai.com/ml-in-production/model-monitoring

Chuong, T. (2016). Evolution of the Netflix data pipeline. Netflix Technology Blog. https://netflixtechblog.com/evolution-of-the-netflix-data-pipeline-da246ca36905.