The entire software's premise is built on some outdated machine learning techniques, techniques that have been refuted in academic literature and research as useless in predicting financial time series.
See that is your issue. You are focused on the machine learning aspect. You say that you have a lot of experience in developing algos for hedge funds, banks etc. So how did you come up with something? You first observed a pattern or you thought of something that may or may not work. You either coded it yourself or the quants did. If there was some merit or perceived edge then off to the developers. This is how I used the program. I never plugged in a bunch of indicators and then hoped and prayed that it found some magical strategy for me that was so convoluted no one could possibly understand. What I did was if I had an idea, such as if the previous day closed down in the bottom 15% of the weeks range and I bought the top 10 weighted stocks of the S&P would it work. If there was nothing, I would move on. If there was something interesting, I would go into it deeper. Whether it was ask a friend to help me with the optimization or backtesting, whatever. It just allowed me to quickly shift through ideas that I had with the point and click interface BEFORE I knew how to code it all up in python and do the research.