Spark streaming application in Jetson Xavier NX

I’m looking for some alternative ways to implement a spark streaming cloud-native app in Jetson Xavier NX.
According to the findings, I know we can use CUDA and on top of it, we can use RAPIDS accelerator for apache Sparks to implement the app.
Apart from CUDA, can we use OpenCL/TensorFlow/PyTorch/scikit-learn frameworks for the same purpose?

Thanks,
Shash