RealityEngines.AI Comes out of Stealth and Launches the World’s First Completely Autonomous AI Service to Address Common Enterprise Use-Cases

RealityEngines.AI, a San Francisco-based AI and machine learning research startup, is coming out of stealth and launching the world’s first completely autonomous cloud AI service to address common enterprise use-cases. The cloud AI service automatically creates, deploys and maintains deep learning systems in production. The engine handles setting up data pipelines, scheduled retraining of models from new data, provisioning high availability online model serving from raw data using a feature store service, and providing explanations for the model’s predictions. The service is the first of its kind and helps organizations with little to no machine learning expertise plug and play state-of-the-art AI into their existing applications and business processes effortlessly.

RealityEngines.AI tackles common enterprise use-cases including user churn predictions, fraud detection, sales lead forecasting, security threat detection, and cloud spend optimization. Customers simply have to pick a use-case that is applicable to them and then point their data to RealityEngines. The service will then process the data, train a model, deploy it in production and maintain the system for them. Behind the scenes, RealityEngines.AI searches several thousand neural net architectures to find the best neural net model based on the use-case and dataset. The underlying neural net model trained by RealityEngines.AI surpasses custom models that are hand-tuned by experts and take months to put into production.

Recommended News: Data Privacy Day: Enhance Data Security in 10 Minutes or Less

Another common barrier to organizations deploying deep learning systems in production is the lack of large volumes of good data. RealityEngines.AI has built on existing research and invented a new technique to effectively handle smaller, incomplete and noisy datasets. The company has developed a technique based on generative models (GANS). This technique creates synthetic data that augments the original dataset. and then trains a deep learning model on the combined dataset. These models yield up to 15% improvement in accuracy compared to models that are trained without using this augmentation technique.

Recommended News: New Data Shows Consumers Prefer Ads Over Paywalls, Giving Publishers a Sustainable Way Forward in Difficult Market

RealityEngines.AI has created a fun visual demo to showcase their technology. The demo generates versions of celebrities expressing different emotions such as happiness, sadness, anger, disgust, and surprise, transforming their age and gender and mouthing famous quotes. The technology works on virtually any photo with a face and can be tested by uploading a selfie. Behind the scenes, a generative model will create multiple versions of the selfie in near real-time. The same technology is re-purposed to create different samples of the original data and a synthetic dataset. This dataset can be used to train robust models even when there is insufficient raw data.

Recommended News: Datadobi Eases NAS Migrations With New DIY Starter Pack