Efficient cloud scheduling is critical for optimizing resource utilization and minimizing job latency, especially in workloads with inter-task dependencies. While existing federated scheduling frameworks like Megha improve scalability, they fall short in handling dependent tasks, leading to performance bottlenecks. In this paper, we propose Fedsort, a dependency-aware scheduling strategy integrated into the Megha architecture. Fedsort introduces task sorting, batching, and prioritized scheduling to efficiently manage inter-task dependencies. We enhance the MeghaSim simulator and evaluate our approach using real-world Alibaba traces and synthetic DAGs. Experimental results show that Fedsort reduces average job delay by up to 76.06.