I really hope/imagine this project specifically has a LLM of some kind doing real-time analysis on the uploaded files for malware from the get-go. How good that is could be is anyone's guess (and chances are there would be blind spots / evasion techniques).