MLOps-Enabled Autonomous AI Agents for End-to-End Software Project Management
International Journal of Innovations in Science and Technology · 22.04.2026 · 614
DOI: 10.33411/ijist/1864 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Innovations in Science and Technology, online erschienen am 22.04.2026. Verfasst von Shawaiz Arif und 1 weiteren Autor:innen. Die Arbeit nennt 27 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The rapid evolution of software project management (SPM) practices necessitates adaptive, intelligent systems capable of managing complex projects efficiently. This paper presents an MLOps-enabled framework of autonomous AI agents for end-to-end software project management, integrating task scheduling, predictive risk assessment, resource optimization, and automated stakeholder communication. The framework leverages multi-agent coordination and continuous MLOps pipelines to ensure reliable, adaptive, and scalable decision-making. Experimental evaluation across 90 simulated and historical software projects within a controlled Discrete Event Simulation environment demonstrates significant improvements, including a 64.98% reduction in average task delays, an 89.06% reduction in critical risk incidents, a 31.8% improvement in resource utilization consistency, and a 67.2% improvement in stakeholder response time compared to conventional Critical Path Method (CPM) and heuristic approaches. Overall Task Completion Efficiency (TCE) reached 92.5% (95% CI: 91.2%–93.8%) and Risk Mitigation Effectiveness (RME) reached 89.3% (95% CI: 87.6%–91.0%) across 50 independent simulation runs. Per-sprint analysis reveals progressive performance improvement from Sprint 1 (TCE: 88.2%; RME: 85.1%; Resource Utilization: 79.3%; Stakeholder Satisfaction Index: 4.2/5.0) to Sprint 5 (TCE: 94.7%; RME: 91.2%; RU: 87.2%; SSI: 4.7/5.0), demonstrating the compounding benefits of continuous MLOps retraining. Inter-agent communication overhead averaged 4.2 ms per message (95% CI: 3.8–4.6 ms), confirming negligible coordination latency. Model drift, measured via Population Stability Index (PSI), decreased from 0.21 to 0.08 across five automated retraining cycles triggered after every fifth simulation sprint, with zero agent downtime during retraining events. These outcomes were achieved through automated model retraining, drift detection, and explainable decision support for human-in-the-loop oversight. These simulation-based results highlight the framework's potential to redefine project management, providing a strong empirical foundation for future real-world validation in enterprise software development environments.
Zitieren
APA
Arif, S., & Ahmed, M. (2026). MLOps-Enabled Autonomous AI Agents for End-to-End Software Project Management. International Journal of Innovations in Science and Technology, 614. https://doi.org/10.33411/ijist/1864
Vancouver
Arif S, Ahmed M. MLOps-Enabled Autonomous AI Agents for End-to-End Software Project Management. International Journal of Innovations in Science and Technology. 2026:614. doi:10.33411/ijist/1864
BibTeX
@article{Arif2026,
author = {Arif, Shawaiz and Ahmed, Muhammad},
title = {MLOps-Enabled Autonomous AI Agents for End-to-End Software Project Management},
journal = {International Journal of Innovations in Science and Technology},
year = {2026},
pages = {614},
doi = {10.33411/ijist/1864}
}
EUVIMED