My for-profit company (Brighter Planet) often gets product ideas from our data scientists; it's exactly what Dr. Haghighi is talking about.

For example: trying to model environmental impact of Bill Gates's 66,000 sq ft house during a hackathon -> discovery that we need fuzzy set analysis (https://github.com/seamusabshere/fuzzy_infer) -> new, marketable capabilities in our hotel modelling product (https://github.com/brighterplanet/lodging/blob/master/lib/lo...).