Chinese start-up DataCanvas wants to democratize artificial intelligence through a plug-and-play platform that lets companies customize algorithms. Financial institutions are the early adopters

When Lei Fang, co-founder and CEO of DataCanvas, came face-to-face with Kyle Liu, a principal at Redpoint China Ventures, in 2016 they discussed the emergence of a data science platform that fused artificial intelligence (AI) and big data. A year later, Redpoint invested RMB50 million in ($7 million) in DataCanvas alongside state-owned Beijing Zhongguancun Development Fund.

Both investors have now re-upped in a RMB120 million Series C round, which also featured Delta Capital, Shidake Investment, and a fund managed by GF Securities.

Since that first conversation, the data science platform has become a reality. The goal is to democratize AI by capturing the knowledge, experience, and best practices of the world’s leading data scientists. Companies can leverage the platform’s machine-learning capabilities to apply algorithms to different aspects of their business, creating customized solutions.

Shanghai Pudong Development Bank, one of the early customers, used DataCanvas to build its big data center. It became the first Chinese lender to monitor and analyze each of its 600 million daily transactions. This is achieved by running more than 50 models simultaneously in real-time that check for abnormal payments and money laundering, coordinate regulatory compliance, and even recommend products to private bank customers.

“This is an upgrade of the whole IT system, from a transaction-based architecture to a cognitive architecture that can identify and analyze each transaction,” Fang tells AVCJ.

Redpoint’s Liu suggested that DataCanvas start out by concentrating purely on financial services and 70% of its customers come from that sector. Rapid uptake by banks, brokers and fund managers has enabled faster product development. The platform has evolved from a rough prototype into a comprehensive and reliable platform.

However, Fang insists that the product remains opensource with all algorithms publicly available. “We open up the code of all algorithms to our customers, so they can customize their own algorithms. We are not recommending a specific or optimized algorithm to customers,” he says.

Fang contrasts this white box stance, with the approach taken by 4Paradigm that offers only selected algorithms. Both Data Canvas and 4Paradigm are identified as majo players in the space by global research and advisory firms IDC and Gartner. 4Paradigm recently raised about $80 million in an extended Series C round.

The white box stance appears to be in line with the regulator’s thinking, given instructions issued last year that all lenders must run risk controls internally, rather than outsource to opaque entities. This impacts a bank’s ability to use 4Paradigm – though it is said to be releasing an opensource product – as well as credit-scoring services provided by the likes of Alibaba Group’s Alipay.

"You can explain the white box results but not the black box results,” Liu says. “Banks and financial institutions must be able to explain themselves. It relates to compliance.”

Having seen revenue growth at more than 100% year-on-year, DataCanvas is now looking to expand into other verticals. Smart manufacturers will be a priority because they have accumulated lots of data that can be used to run visual inspections and product quality controls.

In addition, the company is also lowering the threshold for onboarding customers. A subscription model has been launched as a partial replacement for the existing license model. Customers can start using just one or two models across a handful of terminals, which will bring down upfront payments from the current level of around RMB1.5 million per customer. There is now also a cloudhosted option with initial payments as low as several thousand US dollars.

《Asian Venture Capital Journal》原文链接