Multithreading
Algorithmia can spin up many copies of your Algorithm in parallel, allowing you to break up a large batch job into many parts and have them all run at once.
To make this happen, use whatever multithreading or multiprocessing library you prefer, and create 2-16 parallel threads, each of which will make separate algo.pipe() calls over the data they need to process. This can be done from local code, or placed into a single “master” Algorithm which manages the others.
Here’s a sample Algorithm which takes a single array of inputs, then runs up to four parallel threads at once to call a simple “hello world” Algorithm, reassembling the results into a single output: