Keras, Tensorflow neural nets can run faster

Hi. Would anybody be interested in running a Keras/Tensorflow net 2-5x faster, with less CPU / GPU budget? I've created a C++ solution for my 200+ layer network, which beats the "standard" python inference by far and thinking about making it available for others. I know there are TensorRT and LibTorch, but I think they are a nightmare to use. I'm fishing for early feedback for this project. Contact: marton@cozytap.com