Serverless computing has revolutionized cloud computing by allowing developers to build and deploy applications without the need to manage the underlying infrastructure. However, effectively allocating resources to handle a dynamic workload remains a significant challenge in serverless computing. This paper introduced an approach for autoscaling in serverless environments by combining Deep Q-Learning (DQL) with generative adversarial networks (GANs). The DQL handles dynamic resource allocation in real time based on workload demands. At the same time, the GANs generate synthetic workloads to replicate diverse traffic patterns by training the DQL model to optimize the system’s performance and energy efficiency. The experiments have been performed on the AWS Lambda platform to show the ability of the proposed approach compared to traditional techniques. Our proposed methodology shows improvements in response time, execution cost, throughput, execution time, and number of instances deployed. To address the research gaps in resource management, these findings highlight the option for reinforcement learning techniques in serverless computing.