| This is a great problem space! As mentioned by an other commenter, from the industry point of view it is quitte hard to find the good researchers. I worked on campus (both in Europe and US) for a Major insurance company trying to help with corporate R&D and there have been a lot of hurdle (see it as a list of pain points I faced when trying to collaborate with universities) : - Some researcher are a little too much focus on their area on expertise and will refuse to work on their research outside of the field they have in mind, because either "nobody need it for that, it's mean for [insert previous research project of researcher]". Ex: We need some really precise Geolocation to estimate the level of a river bed for flood estimation near industrial site, found the good lab / researcher (previous paper in the space solving similar problem albeit not for river), go to them ready to propose a grant to work on the topics, not interested to work on river. - Almost no researcher will want to work on a field that is view as 'solved', as they will not be able to publish. Ex: We are desesperate for better OCR tech, as we have an archive of hundreds of millions of document we need to make sense of. No machine vision / CS lab we approach want to do some research in this field as it is solved. Yet in reality the technology is generally not accurate enough on print character. Same story with Handwritten Character Recognition. - Some researcher will try to promote their implementation of their research, but generaly the quality of the implementation is far from what we expect to put in production, and generally have little UX / modern stack understanding, if you are lucky it's implemented in JAVA, but a lot of things are a python notebook, some R or MatLab. They will however not want to collaborate on a industry ready implementation and try to push their implementation they are in love with. If you were to fix this pain points, and provide a tool allowing for discovery this will really be great (and I will be happy to use it). However the really interesting collaboration emerge when you approach a lab for use of their research in a tangent field they didn't anticipate. I wonder if you could automate/accelerate/facilitate this with machine learning or something else. |