Big Data

Skymind launches with open-source, plug-and-play deep learning features for your app

Image Credit: Rosario Lizana/Flickr

In Silicon Valley, deep learning ranks as one of the hottest technologies. Now, this startup sees a chance to let lots of developers incorporate deep learning into their apps.

Deep learning essentially drives computers to make inferences like human brains do, based on the consideration of lots of information. It can be used to identify patterns and other traits in images, music, and text.

A 24-year-old named Adam Gibson came up with open-source software called DeepLearning4j to do all those things; and today, Gibson is launching Skymind, a startup that will provide commercial support for the use of that software.

San Francisco-based Skymind already has two paying customers, and the company intends to help lots more companies in the deep-learning game in the future.

“Rather than [have] this crazy mission, no, it’s just enablement and practicality,” Gibson said in an interview with VentureBeat.

With deep-learning projects from tech giants like Baidu, Google, Microsoft, and Netflix getting so much attention, it’s not too hard for tech people to get curious about deep learning and wonder what it would take to help more companies use it.

Investors seem intrigued. A startup focusing on hardware for deep learning, Nervana, picked up $600,000 in seed funding recently.

Skymind has chosen a different route for running a business around deep learning. Its model resembles Red Hat, in the sense that it’s backing up open-source software with commercial support.

It turns out other open-source software for deep learning exists. Lots, like PyBrain and Theano, are based on the Python language. But Deeplearning4j, in Java, could be more broadly useful. Java is core to the Hadoop open-source software for storing and analyzing lots of different kinds of data, and if Skymind’s software takes off, Java could become more widely used for deep learning, too.

Skymind’s roadmap is extensive. Plans call for the construction of a graphical user interface that could help Deeplearning4j become easier for lots of people to use, as well as a software-development kit compatible with the jQuery JavaScript library. Skymind also wants to make the software compatible with graphics cards from companies like Nvidia.

But it’s a matter of balancing commitments. Gibson currently teaches machine learning at the Zipfian Academy data science training program in San Francisco.

In fact, Zipfian has partnered with Skymind to train people to do deep learning, Gibson said.


VentureBeat is studying mobile marketing automation. Chime in, and we’ll share the data.
0 comments