H2O.ai confirms its status of leader

During one of the best meeting of the recent IT Press Tour with H2O.ai, its CEO and founder Sri Satish Ambati, gave us a good intro and state of the art of AI.

Founded in 2011, the company has raised so far $73.6M in 5 rounds to democratize Artificial Intelligence (AI) via its open source Machine Learning (ML) platform. The idea is to offer a foundation to build or improve applications illustrating the various use cases identified by H2O, critical for the market take off.

Driveless AI is a big iteration in AI adoption for enterprise, it automates data science and machine learning workflows with a high level of accuracy. The impact is impressive with an unique UI for immediate visualizations.

The solution runs on commodity hardware specifically configured with GPU for parallel processing.

ML automation solves also the challenge of data scientist limitations helping companies to boost workflow, visualization and modeling. Driverless AI accepts data from different sources such tabular data from plain test files (csv, txt, data, orc, svmlight, avro, parquet, tgz, gz, bgz, zip, xz, xls, xlsx, bin and arff), Hadoop, Google cloud storage, Google BigQuery or even S3 buckets. Following the data import, the dataset is ready for an experiment and can be monitored in live mode with a very unique UI.

In details, a data is a row representing one entity and must have an associated label, all data present in one dataset. It also means that data submitted to the ML engine is prepared for that with cleaning and normalization already done.

Driverless AI reduces drastically the data processing time with automation and optimized data modeling implementation. It delivers a better prediction model, more easy to deploy and reuse with standards such Python and Java-based technologies.

Technically for on-premises deployment, Driverless AI, installed as a Docker image, requires a minimum of 10GB of free disk space to run and is greedy of memory with a minimum of 64GB dedicated to its job and at least 10x of your dataset size of disk space. For cloud installation, I should say for Amazon, H2O provides a AWS AMI. On bare-metal, a RPM package is available for Red Hat, Centos 7, SLES 12 Linux OS, a DEB is also available for Ubuntu 16.04 and for Windows with Subsystem for Linux (WSL).

H2O is a real breakthrough for current enterprise use cases with a proven tool more and more adopted by the market. If you have a chance to attend an H2O conference, soon in London and in a few months in San Francisco, join and listen carefully, something is happening...