A startup called Algorithmia Inc. is pitching a new software platform for developers who don’t have the specialized infrastructure needed to run artificial intelligence technologies.

The company said its new Algorithmia AI Layer can accelerate the deployment of AI and machine learning models inside software applications. Designed for companies that want to build large AI and machine learning portfolios, the software also lets developers automate their software development processes and deploy models written in their preferred programming language, via an application programming interface, in a matter of minutes rather than weeks.

For the uninitiated, AI and machine learning models are defined as the mathematical models that are created when algorithms are trained using a particular data set. The algorithms are trained using data that contains the correct answer to a problem (known as the “target”), as well as many incorrect answers. Once the algorithm “learns” the answer to the problem, a model is created that can be used to make inferences on new data that doesn’t contain a target.

“[The] models are usually modeled around past behavior in an effort to predict a future classification,” Algorithmia’s chief executive officer Diego Oppenheimer said. “This could be anything from looking for anomalies in a time series to finding objects in an image.”

Oppenheimer says automated deployment is necessary for the models because organizing and implementing these manually is an extremely complex process. Deploying them can often take weeks, and developers also need to put in many hours of work to keep them up to date. The main issue is that developers are constrained by a lack of AI infrastructure that Algorithmia says has become the “last mile” problem for developers, and keeps many AI projects from reaching production.

“Tensorflow is open-source, but scaling it is not,” Kenny Daniel, co-founder and chief technology officer of Algorithmia, said in a statement. “Almost all R&D has focused on collecting and cleaning data, and building models. Algorithmia has spent the last five years building the infrastructure that will put those models to work.”

Algorithmia reckons the best solution to these developer problems is a dedicated AI Layer, which acts as a host to these models and slots inside companies’ infrastructure stacks.

The company is providing two options for developers, namely a serverless and an enterprise version of its AI Layer. The serverless edition is basically a cloud-based version accessed via the company’s website. Users simply log in to the platform, choose the models they desire and then host these on Algorithmia’s cloud via an API.

The enterprise AI Layer, meanwhile, can be deployed on public and private clouds, with models added via Git and utilized with an API call.

“Anybody can stand up a simple web server, but not everyone can run Reddit,” Oppenheimer said. “A similar analogy applies to AI and ML. I can fairly easily put an API around an ML model, but scaling it so that hundreds of thousands of users on a website or mobile apps can call it in a secure, reliable and auditable way is a completely different story.”

Although it’s a new platform, Algorithmia’s AI Layer has already been endorsed by Google LLC’s artificial intelligence experts. The endorsement comes after the Internet giant’s AI and machine learning fund participated in Algorithmia’s $10.5 million Series A round of funding in June.

“As someone who has spent years designing and deploying machine learning systems, I’m impressed by Algorithmia’s serverless microservice architecture – it’s a great solution for organizations that want to deploy AI at any scale,” said Anna Patterson, vice president of engineering for artificial intelligence at Google.

Algorithmia said its new AI Layer is available now. The serverless edition is being offered on a pay-per-use basis, while the enterprise version is sold as a traditional enterprise license based on the number of compute nodes used.

Image: Algorithmia