We believe deep learning and AI are going to fundamentally alter the world. And we want to be right in front of that change as it happens.

We founded Agara Labs as a vehicle to develop and deploy deep learning-powered solutions for problems that enterprises are faced with. The solutions need to be expansive, covering entire verticals at one go, rather than focusing on one narrow, specific issue. The solutions need to be applicable right away, not in the distant future. The solutions need to simplify matters, not add a layer of confusion over the matrix of solutions enterprises typically work with. Finally, the solutions need to be delivered in the simplest and most elegant of interfaces to ensure we are empowering people, not weakening them.

In the last few months, we have been hard at work developing solutions for customer service — that all-important vertical that makes or breaks companies.

The product we are developing does two very large tasks — multi-parameter query classification and construction of conditional, different-sized responses. Hidden beneath these esoteric terms is years of training and validation data, multiple models, relentless hyperparameter tuning and innovative pre and post-processing of data. All of this is being delivered to customer service professionals in an effective user interface with very little training requirement.

Importantly, we are not alone in this journey. We are working with a few large global enterprise customers who share our vision of what deep learning can achieve. They are sharing with us both their product expertise as well as reams of real-world, near-realtime data. Our products will soon be going to production serving their customer care vertical. They get access to cutting edge technology and we get willing partners to test and perfect the technology.

Over the next few months, we will be adding more capabilities of which one is particularly interesting. We envision the product to be able to decide the appropriate course of action from among multiple (sometimes complementary, sometimes conflicting and always confusing) choices available to a person. This will require heavy-duty application of cutting edge deep learning technologies. We can’t wait to get to that phase.