Machine Learning Operations (MLOps) has become increasingly essential for research labs aiming to streamline machine learning workflows and ensure the reproducibility, efficiency, and scalability of their models. However, many research labs face significant challenges due to limited hardware resources, which hinder their ability to implement robust and scalable MLOps systems. This paper aims to address these challenges by proposing scalable MLOps architectures specifically designed for research environments with constrained hardware capabilities. We have proposed a scalable MLOps architecture specifically designed for research labs with limited hardware resources. Our results show that the proposed architecture improves resource utilization, offering a more resource-efficient solution for research labs with limited hardware. Furthermore, we provide illustrative cases of real-world deployments and the benefits of this solution in practical research environments. The main contribution of this work lies in its ability to provide a practical and scalable solution for research labs operating under hardware constraints, ensuring they can continue conducting advanced research without being limited by computational capabilities. The solutions presented in this paper not only highlight the efficiency of the proposed MLOps architecture but also lay the foundation for future studies to further optimize workflows in resource-constrained environments.