I wrote a "resource forecasting and employee scheduling" system that used what would now be called deep learning (I did this in 2002) that consumed multiple yea

I wrote a "resource forecasting and employee scheduling" system that used what would now be called deep learning (I did this in 2002) that consumed multiple years of an animation studio's accounting records, covering everything the studio spends money, including fine grained information about the employees, what jobs they worked on, their specific roles, who they helped accomplish tasks, and tasks they accomplish alone. This system operated in Excel, as that was the tool the studio execs were used to using, but I had to write the entire thing in C++ and access it as a dll within Excel because it referenced gigs of data not in spreadsheet cells. The studio I worked, Rhythm & Hues, was a 2-times Academy award winner. My "resource forecasting and employee scheduling" system was able to prove the presence of institutional learning, and the prediction graph of that learning curve was used by banks to justify the working capital loans made to the studio. It also maintained the forecast of the render farm usage, identifying when and how many additional render servers needed to be leased. Perhaps the most useful was it generated on-demand employee schedules: who works on what, who is most likely to be helping, and insures those that support one another most frequently are scheduled with compatible tasks. The studio owner thought I should market and sell that system independently. I also came up with a VFX process for automated actor replacement, which I did purse - trying to create Personalized Advertising with you and your friends in the video ads you see when streaming. However, VCs wanted to finance a porno company (seriously! they insisted) and I eventually went bankrupt. But hey, I can claim high morals as the global patent author that invented deep fakes, but refused to do porno with it.