I have a Python-based maximum entropy classifier. It's large, stored as a Pickle, and takes about a minute to unserialize. It's also not thread safe. However, it runs fast and can classify a sample (a simple Python dictionary) in a few milliseconds.

I'd like to create a basic Django web app, so users can submit samples to classify in realtime. How would I load the classifier into persistent memory once, and then regulate it so that each request can access the object without conflicting with other requests?

you could use djangos cache-framework and set the timeout to a extreme value