Addressing the challenges of managing high-performance computing workloads in dynamic cloud environments, this paper presents a SLURM-based reference architecture. We elaborated Infrastructure as Code (IaC) to automate the deployment and management of SLURM, enabling efficient resource allocation and scalability. The basic scheduler architecture descriptor was further extended with computational tools and frameworks required by the HUN-REN Cloud scientific community. Results from benchmark experiments show significant performance improvement through parallelization, demonstrating SLURM's ability to utilize cloud resources for fair workload management of calculationheavy tasks. Our AlphaFold protein structure prediction experiments demonstrate an 82.1% reduction in computational runtime when scaling from 1 to 8 worker nodes, with execution time decreasing from 3154.75 seconds to 563.75 seconds.