The use of high-performance computing techniques is mandatory for exploration geophysics. That is because of the ever-increasing amount of data and the complexity of the algorithms. Given that, we have developed research on improving the parallel efficiency of such applications. Our previous work presented auto-tuning and load balancing techniques for wave-based methods such as acoustic modeling and reverse time migration (RTM). We also developed an automatic generation of highly optimized finite-difference GPU kernels from high-level symbolic equations by integrating Devito and OPS. We are currently further exploring the load balancing of geophysical methods, especially the full waveform inversion (FWI). We are also active contributors to Devito's project, where we mainly work on enabling efficient parallel finite-difference kernels on multiple GPUs.