Customers include Google, Intel, and the US government
California-based Snorkel AI, a startup spun out of the Stanford AI Lab, has raised $15 million in funding as it emerges out of stealth mode.
The funding for Snorkel’s end-to-end machine learning platform came from Greylock, GV, and In-Q-Tel (the investment arm of the Central Intelligence Agency).
The startup’s customers include US banks, large enterprises, and government agencies.
Focus on training data
Snorkel focuses on large training datasets, programmatically labeling and managing the training data that fuels AI models.
For example, banks use the Snorkel Flow platform to build AI applications that classify and extract information from their loan portfolios.
While at the Stanford AI Lab, the Snorkel founders developed technologies for training data that were used by companies including Google, Apple, and Intel.
“Despite spending billions of dollars on AI, few organizations have been able to use it as widely and effectively as they want to,” said Alex Ratner, CEO of Snorkel AI.
“Available solutions either ignore the most important part of AI today – the labeled training data that fuels modern approaches – or rely on armies of human labelwers to produce it. Snorkel Flow focuses on a new programmatic approach to the training data that enables enterprises to use AI where they couldn’t before.”
Developers using the Snorkel Flow platform create ‘labeling functions,’ or rules, and other programmatic operators, which the platform automatically integrates to train machine learning models.
Users then can then adapt the models by editing the programmatic training in Snorkel Flow’s interface.
“We’ve consistently heard from Fortune 500 CIOs that they have been disappointed with their progress using AI, largely because they get stuck on the data,” said Saam Motamedi, partner at Greylock and Snorkel board member. “Customers’ rapid success with Snorkel Flow is a testament to the power of this new, data-centric approach, which has the potential to democratize AI across the enterprise.”
Snorkel Flow has been used for a number of applications. For example, Google used the platform to replace hand-annotated labels in key machine learning pipelines, Intel used it to replace a high-cost, high-latency crowdsourcing pipeline to accelerate sales and marketing agents, and researchers at Stanford Medicine used Snorkel Flow to label medical imaging and monitoring datasets, replacing person-years of hand labeling.
“The time, expertise, and costs involved in labeling training data present significant challenges to the US government in applying AI to missions of national security,” said A.J. Bertone, a partner at In-Q-Tel. “Snorkel AI provides a revolutionary capability that can greatly reduce the level of effort required to develop mission-ready machine learning models by addressing this critical data problem.”