We're running a cloud-based HR platform for companies in Pakistan. The payroll module handles thousands of employees across hundreds of clients. Month-end batch processing is getting slower as we scale.
Currently, we're using CPU-based processing (Intel Xeon) with Java microservices. The bottleneck is the sheer number of calculations - tax slabs, social security contributions (EOBI, SESSI, PESSI), and company-specific deductions. It's a lot of math.
The question: Has anyone used GPU acceleration for this type of batch processing? We're looking at:
- Parallelizing employee calculations (each employee is independent)
- Offloading math-heavy operations to GPU
- Using CUDA or OpenCL for compute
We're using Vertex HCM for the platform side and wondering if GPU compute would help with the backend processing.
Anyone tried this approach for similar data-heavy applications? Any libraries or frameworks you'd recommend?
