Scaled Inference Lands $8M From Khosla To Build A Cloud-Based AI Platform For All

Next Story

A Gift Guide For The Young At Heart

Scaled Inference — a startup founded by two ex-Googlers that is building a cloud-based platform for third parties that want to use artificial intelligence and machine learning tools to run their apps and services — has raised another round of funding to continue its development and hiring, a Series A round of $8 million from Khosla Ventures.

This brings the total amount of funding up to $13.6 million, after the startup raised a seed round earlier this year from a number of angels and prominent investors like Tencent and Felicis.

While Scaled Inference has yet to launch a public product, it expects to have a working prototype together soon and is now bringing together companies for a closed trial of the platform.

In the months between raising its seed and Series A rounds, Scaled Inference has been putting together some of the first products (or services) that will run on top of that platform: pattern recognition, anomaly detection, prediction, and predictive ranking, which will be accessible by developers by way of a set of APIs.

The challenge for Scaled Inference is that it is not only building a cloud platform for machine learning intelligence, but also trying to figure out how to distil what have up to now been very specific applications into general-purpose services. This first set, says co-founder Olcan Sercinoglu (who started the business with Dmitry Lepikhin), will “serve as a foundation for various special-purpose services that we also plan to offer in the future.”

“These core services all depend on a user-supplied data set or stream that largely defines the potential applications,” he explains.

For now, a lot of the services are focused around e-commerce and related financial services.

As an example of where pattern recognition might be applied, he says, take a bank or credit card company like Visa. Visa would feed transaction data like time, location, merchant, customer, item, and amount for each transaction.

“The pattern recognition APIs can then be used to detect correlations such as increased spending during certain time periods, at certain locations, by certain customers, on certain items, for certain combinations of these attributes,” he says. In turn, this type of insight can help guide important business decisions such as branch locations/hours, regional marketing strategies, pricing for transaction fees, and so on.

Anomaly detection APIs, meanwhile, could be used to detect unexpected events like a large transaction at a coffee shop, or to single out transactions made in a geographic area where a customer is not usually based. Of course, these are the types of services that large banks (or Visa) would already have in place, but now the idea is to create these services in such a way so that small financial startups, for example, might be able to execute similar anti-fraud data inquiries.

The prediction APIs focus on predicting attributes based on other attributes. One example is the amount that will be spent in a particular region during a certain time period. “This type of precise predictive analysis can be used to optimize important future-looking business decisions, or to enable intelligent products or services that can save users time by predicting what users will want or need before they ask for it explicitly,” he says.

The predictive ranking API is related to this, and will let users rank large numbers of objects based on predicted attributes. “For example, it could be used to get a ranking of items that a customer is most likely to purchase at a certain time/location, or alternatively a ranking of potential customers for a given item or merchant,” Sercinoglu says. In addition to shopping apps, these could also be used for advertising systems sorting through inventory.

There is a potential roadblock for Scaled Inference applying these services to e-commerce companies purely in the public cloud: these types of applications are based on sensitive personal data, such as financial or health data, and as such may require private or on-premise deployment of its technology due to government regulations.

In that regard, this is a long-term vision, but this is one of the reasons that Scaled Inference went with the investors that it has done, Sercinoglu says. “Vinod Khosla is a long term investor, one of the longest-term thinkers in the business, and ours is a long term vision for AI and machine learning. That really resonated with him.”

Longer term, he says, “We seek to offer the same powerful technology as a public cloud API to everyone and for virtually any application; anything from personalized magazines to intelligent context-sensitive address books, app launchers, e-commerce sites, disruptive data-driven startups (Uber, Airbnb, Rent a Runway, etc), advertising systems, search engines, and so on. Ultimately, our goal is to power the most intelligent and popular products and services in the world.”

Featured Image: Marius B/Flickr UNDER A CC BY 2.0 LICENSE