The rapid growth of data in modern databases has made it increasingly difficult to retrieve accurate information, particularly when dealing with spelling errors or linguistic variations. This paper addresses developing an efficient search suggestion generator using the Double Metaphone algorithm, a phonetic encoding technique that converts words into codes representing their pronunciation. This procedure consists of two phases. In phase 1, it will generate phonetic codes for searchable words, and then in phase 2, it uses those codes to match against queries from users and returns suitable matches. Here, the system is parallelized, employing the Message Passing Interface (MPI) in distributed environment and Compute Unified Device Architecture (CUDA) in GPU-based parallelization to speed up the processing of very large datasets. The division of data among multiple processors and/or the CUDA threads to be processed reduces query response time and enables scalable retrieval. This paper compares the performance of both the sequential and parallel approaches showing the effectiveness of phonetic encoding and parallel computing to improve search accuracy and efficiency for large databases. The experimental results show that the processing speed for CUDA is much greater than MPI, while MPI was able to yield an improvement greater than the sequential implementation. This indicates that a combination of Double Metaphone algorithm with parallel computing techniques does yield an extremely efficient and scalable solution for high-performance search systems.