‘To me, it was very European. You know, it was nine o'clock at night, lots of people out walking and jogging by the waterside.’ As well as these high-altitude global partnerships, it is notable that Ubotica has deep roots in computer vision research in Ireland, through university labs and an earlier company called Movidius. Ubotica co-founder, John Bourke, transitioned directly from Movidius to Ubotica. Movidius was acquired by Intel in 2016. ‘That was the genesis for the founding of our company, and we were very fortunate. My co-founder, Aubrey Dunne, also worked in Movidius, but he was a contractor, and Intel had a policy that they did not hire contractors, so Intel's loss was our gain. After completing their tenure with Intel following the acquisition, the other two founders of Movidius – Sean Mitchell, and David Moloney – joined us.' With the Ubotica co-founders, the chairman and CCO (Mitchell) and chief scientist (Moloney) primarily coming from Movidius, along with other employees, the groundwork laid by that company, going back to the mid-2000s, can hardly be overstated. The improvement of machine vision standards by Irish researchers has applications in terrestrial industries of immense value to the republic: intelligent eyes must pick out potential faults in agricultural products (established, in use), in pharmaceutical (emerging, difficult) and in silicon wafers (highly complex). As with the evolution of eyes in living creatures, each advance in computer vision has led to upsurge in adoption, as industry players press to remain competitive. Jonathan Byrne, a scientist who worked at Movidius and is now in charge of automation and Big Data analysis at the Intel fab in Leixlip, added that, ‘Machine vision enables a new level of automatic decision making that was not previously possible – but is now required to stay productive and drive down costs.’ How does Buckley see this technology cluster progressing? With combining AI+computer vision for pharmaceuticals, for example, he sees the barriers as tricky, with the regulations and risks involved in healthcare mean it’s hard to ‘fully trust’ the machine to act. With space tech, the question may be whether it is necessary to source their own hardware at all. ‘We are evolving into being hardware agnostic. So we don't care whether it's our hardware that's on the satellite or someone else's. What we are implementing is the glue that pulls everything together in terms of how you deploy the models, how you train them, because it's a challenge when you're running algorithms on board a satellite that you can't touch.’ To train an AI model, you need images. As new sensors are being refined and developed, how can Ubotica train their AI models to be hardcoded and effective before the satellites fly? This question is perhaps where the venture shows its greatest bite-strength. ‘What we have demonstrated is that we can take an AI model, train it on generic data that's not targeted at a particular sensor, and we can do, say, vessel detection with image data onboard the satellite, take that analysis and get that down to the ground, all within five minutes.’ With their concept proven and SpaceX satellites a-swarm around the earth, the mission for Ubotica, headquartered on the Old Finglas Road in Glasnevin, seems clear: to set up the most attractive stall in this highly specialised niche before someone else does.
Download PDF file