Hi -- I work on the fraud team at Stripe. Last October we introduced Stripe Radar ( https://stripe.com/radar ) a suite of integrated tools to help you beat fraud. While different businesses see different patterns of fraud, we believe that overall our fraud prevention is now as good or better than essentially every other product in the industry. This partly follows from Stripe’s scale: whenever a card is used, there’s now an 80+% chance that we’ve seen it before. We use this data from the global Stripe network to detect and prevent fraud with ML models that we’re constantly updating and refining.
We’d love to hear what would make us work better for you. My email is brahn at stripe.com.
Ah machine learning.. I've studying that the last few months. Cant say i got much more wiser on it, as 99% of the examples and explanation out there is regarding classifying images. And not so much about handling structured data. Is there are public information on how ML is implemented at stripe?
And you write that > 80% of the worlds issued cards are in the stripe db because they are once used via stripe? Are you sure about this number?
Can't say im much impressed by this sql interface 'feature'; It's more opening low level access then a feature, but that is just my humble opinion. And it is of course nice for techies having to work with, instead of specific 3rd party APIs.
Will there also be plain (not plaintext but simple) replication interfaces/APIs in the future that clients can run sql slave nodes (mysql/postgresql) with near realtime replication for local querying (privacy) and backup purposes? (and being able to continue to use the data even if stripe would be unaccessible/down)?
We’d love to hear what would make us work better for you. My email is brahn at stripe.com.