Optimizing AI development efficiency through MLOps and security integration: A measurement approach using data envelopment analysis (DEA)

The growing use of artificial intelligence (AI) in engineering industries is exacerbating the trade-off between speed of development and security. Although MLOps help with automation and scaling, they are often perceived as introducing restrictions to workflow that slow down AI efforts. There is hardly any quantitative evidence regarding the impact of security integration on the development efficiency.In order to quantify the system-level evaluation framework of MLOps and security practices (MLSecOps), researchers assess efficiency in AI development using Data Envelopment Analysis (DEA). Eight AI projects that were measured as Decision Making Units (DMUs) evaluated under standard workflows and integrated MLSecOps pipelines. The DEA model takes into account six input variables, such as development time, training time, dataset volume, infrastructure scale, computational resources, workforce effort, and three output variables, namely performance, robustness, and speed of operation.According to constant and variable returns to scale results, all of the MLSecOps pipelines reach full efficiency score while traditional workflows have significantly lower efficiency scores. According to Slack analysis, automation and ongoing security integration can decrease labor effort by 30–40% and development time by around 25%. The results show that embedding security in automated MLOps (MLSecOps) pipelines helps improve efficiency, robustness, and scalability without impacting performance.