US-based digital asset exchange Coinbase has revealed that it’s focused on building “state-of-the-art” machine learning (ML) technology with “efficient” execution for the crypto and blockchain-focused economy.
Catalin Tiseanu, an engineer working at Coinbase, writes in a blog post that the digital assets firm leverages machine learning or ML for several different use-cases, including fraud prevention, and keeping its users safe. ML is also used by Coinbase to curate and personalize content or determine whether identities are matching (or not).
Tiseanu explains that one of the main differences between ML initiatives and other software engineering projects is that it’s difficult to forecast or accurately predict the impact of a particular machine learning model before trying a prototype out. Tiseanu, who previously worked as a Big Data architect at Bitdefender, notes that this is why “getting to an initial prototype and having a fast iteration cycle is critical for machine learning projects.”
He adds:
“With these [design goals] in mind, [Coinbase] set out to build a suite of machine learning models which balance the tradeoff between quick iteration cycles and incorporating new state-of-the-art practices.”
Tiseanu confirms that Coinbase has developed several high-performing ML models, including EasyML, Seq2Win and EasyBlend.
EasyML allows developers to create models on tabular data, e.g. data which includes several different features, like numerical (user age on site), categorical (user city) or dates (last transaction date) — “spreadsheet data.” Meanwhile, Seq2Win enables creating models which use sequential data, like sequences of front-end user events, Toiseanu notes.
He also mentions that EasyBlend enables the exchange to combine several different models together. He claims that “taken together, these 3 frameworks can be used to build machine learning models for brand new use cases, in just a few hours.”
Tiseanu also states:
“By having these building blocks, ML engineers at Coinbase can quickly build state of the art models, without sacrificing either performance or iteration speed. This has led to meaningful business impact, higher productivity by ML engineers, and faster iteration times. Going forward, we’re excited about improving on these building blocks with pre-training Seq2Win and using Graph Neural Networks for handling blockchain data.”
(Note: for more details on how Coinbase is leveraging the latest ML tech, check here.)
As covered in November 2020, Fintech solutions offered by banks might not be effectively using AI and ML due to the lack of qualified professionals, according to a recent report.
Nasdaq CEO Adena Friedman has stated that ML, blockchain, and Cloud technologies will “drive evolution” of capital markets. As reported in October 2020, Mastercard confirmed that it has been using AI and machine learning solutions so that it can prevent fraudulent activities which have increased significantly due to COVID.