In a world the place Twilio exists, you wouldn’t dream of standing up your individual SMS messaging stack throughout 193 nations and god-knows what number of telcom operators. The state of affairs for machine studying (ML) isn’t totally dissimilar; except ML is core to your software program — and it in all probability isn’t — why would you waste time on assembling a whole infrastructure. To resolve that exact situation, Slai is constructing a developer-first platform for machine studying. It equips builders with the instruments to shortly ship machine-learning purposes.
“At present, machine studying stays a analysis self-discipline, and it’s nonetheless very onerous for a developer to construct their very own machine studying utility,” shares Eli Mernit, co-founder and CEO at Slai. “Our hope is that builders are empowered to construct state-of-the-art machine studying fashions.”
The corporate at present introduced it raised a $3.5 million seed spherical led by Tiger International, with further funding from Y Combinator, Cost Ventures, Uncorrelated Ventures, Twenty Two Ventures and Soma Capital, together with angels together with Man Podjarny and Jason Warner.
The corporate’s product is concentrated on letting builders give attention to the machine studying fashions, slightly than on all the encompassing kerfuffle that takes up a variety of time, however doesn’t immediately contribute to the appliance itself.
“The product enables you to join a knowledge supply. That may very well be your database or an S3 bucket with knowledge that you simply wish to ship to a machine studying mannequin. After which the machine studying mannequin — just a few Python code — finds predictions within the knowledge. We’ve wrapped that in an API, that does issues like validation on the enter that the consumer passes in, or does some processing on the output earlier than it’s despatched again to the consumer,” explains Mernit. “These parts represent a machine studying utility. And so usually, if somebody was doing these things by hand, they must arrange an online server themself. They must arrange some versioning system, they must arrange a way of monitoring the mannequin. And all of this quantities to a variety of busywork. We do all that for the consumer. All they need to give attention to is the place their knowledge is coming from and what kind of mannequin are they utilizing. The remaining is dealt with for them. In a nutshell, we eradicate all of the glue that goes into the machine studying growth course of.”
The platform thinks of itself as GitHub for ML — and makes it straightforward to fork current recipes for machine studying to be used in purposes.