The exponential growth of Big Data has exposed significant memory bandwidth limitations in traditional CPU-based DBMS. While GPUs offer a high-bandwidth alternative for analytical processing, current state-of-the-art solutions frequently rely on proprietary frameworks like CUDA, resulting in vendor lock-in and reduced hardware portability. Furthermore, effectively orchestrating query execution across heterogeneous environments remains a complex challenge involving data transfer bottlenecks and memory capacity constraints. This thesis introduces SYCLDB, a cross-vendor, heterogeneous analytical DBMS built upon the SYCL open standard. Extending a previous kernel-library prototype, this work integrates Apache Calcite and Apache Thrift to provide a complete SQL parsing, optimization, and communication layer.