This research introduces a dynamic task offloading mechanism to optimize resource allocation in hybrid cloud-edge architectures for multi-tenant environments. Tasks are intelligently classified and offloaded based on latency and computational needs, leveraging real-time monitoring and machine learning for efficient resource distribution. Reinforcement learning enhances task allocation over time, achieving significant performance improvements. Experimental results reveal a 30% latency reduction and 25% better resource utilization, with validation conducted on a hybrid testbed using simulated workloads. The framework is particularly effective for latency-sensitive applications like real-time video streaming and large-scale data processing, demonstrating a 20 ms latency reduction and 15% throughput increase. Despite its success, challenges such as scalability and energy efficiency remain, warranting further exploration, especially for large-scale IoT deployments. This study highlights the framework’s potential for advancing hybrid cloud-edge solutions while addressing emerging demands in resource optimization.