Snorkel.AI 17 pipeline is what helped transform Snorkel from a research project to the commercial platform it is today. But with such a busy AI space, how does Snorkel stand out from the competition? “As we approached the end of our PhDs, we knew we had an extremely robust paradigm , and there could be a real opportunity to do this in a first- class citizen, enterprise support level way,” Braden recalls. “Now, we have the benefit of being a relatively young company with a mature technology in Silicon Valley terms.” Snorkel came into the field with its credibility thoroughly vetted. The technology had broken free of the lab and was already working on complex real-world problems. Snorkel Flow had been successfully deployed internally at Google and a co-authored paper had been published. It had also been used by Stanford Medicine , Intel , and the US Department of Justice . It was clear that it was a broadly applicable technology that could work across a range of different domains and with any of the machine learning architectures available. The technology is based on new insights into probabilistic graphical models and is inspired by work on bootstrapping and crowdsourcing . There have been 39 publications about the algorithm itself, applying it across modalities from text extraction to images, time series, and video data.