The rapid growth of real-time sensor applications in edge environments requires efficient data processing under strict resource constraints. This paper proposes a TinyML-based framework that integrates lightweight model design with a multiobjective optimization strategy to balance accuracy, latency, and computational cost. The framework is evaluated through a Matlab-based edge simulation using multiple TinyML configurations. Experimental results show that the optimized model achieves an inference latency of approximately 15 ms and a memory footprint of 140 KB, significantly outperforming the baseline model. Moreover, the computational cost is reduced to the order of 105 FLOPs, indicating improved energy efficiency. Despite these reductions, the proposed approach maintains a competitive accuracy of around 94%. The results demonstrate that the proposed framework provides an effective trade-off between performance and efficiency for real-time sensor data processing.