Congrats Polarr! I interviewed with Borui and team back in 2014, when they were still crashing their investor's offices. I was super impressed by the founding team and all their creations since then. Wishing them the best.
If you search CoreML, TFlite or Caffe2 on GitHub, there are already a score of offline AI models on the vertical of image recognition and vision. While these guys might have gotten lucky landing a couple deals with phone makers to enable some 'AI', it's hard to imagine the real AI leaders like Apple, Google or FB making models more and more accessible in the future.
It can still take a few hours to put together or days to test performance. Apple's thermal issues could also make development difficult. This is an area we're working on to ensure stability of our modules running across multiple iOS devices.