two years ago ,I have rewrite some mleap code to support pytorch tensor convert spark dataframe ,Now I want to contribute these code in future, and now for flink cdc real time etl video data ,we will load pytorch javacpp model on flink,so I think mleap maybe support flink and pytorch javacpp write in scala .what do you think