Is it possible to use RAY to distribute a graph-based computation task written with NetworkX API? For instance, computing the page ranks for a big graph.
If the problem is amenable to parallelization, you can use Ray to do it yes (Ray is compatible with networkx and you can pass networkx graphs into Ray tasks/actors).
If you want to compute a page rank, it will probably be more efficient to use sparse matrices. There is some pretty old code in GitHub - ray-project/raynomics: Experimental genomics algorithms in Ray that might help you getting started on that if you are interested!