Dataiku 4.1 GA
November 14, 2017
a wide range of new and improved features aimed at making data science,
machine learning, and advanced analytics accessible to organizations as
a whole, data science software maker Dataiku has released Dataiku 4.1.
The new and improved software platform acts as a central hub for
technical and non-technical users to prototype, build, scale, deploy,
and manage advanced data science products.
"We are focused on building a platform that is a single hub for an
enterprise's data science and machine learning development - that's what
this release reflects," said Florian Douetteau, CEO of Dataiku. "Many of
our customers already use Dataiku with hundreds of users from all
different backgrounds, from data engineers, to developers, to
non-technical analysts, to perform advanced analytics and to develop
data science solutions. This latest release strengthens enterprise scale
development and deployment of these solutions among and across teams."
A new Dataiku with the same great functionalities, built for scale
Building upon the needs of Dataiku customers who have hundreds of users
across their organizations around the world relying on the software,
Dataiku 4.1 has been designed to accelerate scalable deployment while
maintaining its powerful core functionalities such as:
•point-and-click interfaces for data
preparation and analysis
•customizable tools to facilitate
cutting-edge and efficient data science
•straightforward solutions for
deploying, monitoring, and governing models in production.
In its latest release, Dataiku is introducing new features that further
expand its capabilities as a single platform for everyone, including
coders and clickers, spread across any sized organization around the
world. "This release plays to our strength of enabling our largest
customers to propagate data science expertise throughout the
organization," said Douetteau. "In fact, organizations who deploy
Dataiku at scale have on average a 4:1 ratio of non-coding data
specialists to data scientists using Dataiku."
Data preparation tools for coders and non-coders
Dataiku 4.1 introduces new data preparation "recipes" within the Dataiku
graphical interface that bring powerful analytical functionalities to
non-coders, including pivoting, sorting, and splitting datasets.
For coders, the latest release brings advanced visualization libraries
like RShiny and Bokeh for rapidly creating engaging interactive web
applications within dashboards. Additionally, RMarkdown reports let
users easily share their results outside of Dataiku.
Live Model Competitions - Compare models in real-time
With Dataiku's "live model competition," users compare the performance
of a batch of machine learning models competing in real time without
waiting for the entire training of the model. This reduces the training
time and resources used by interrupting or resuming the competition once
it yields satisfactory insights.
Additionally, model ensembling, which exploits the strengths of various
models by combining different algorithms, is now possible without
writing a single line of code.
coding environments for project stability
It is common for an organization to have many projects using different
versions of Python, R, and libraries. Dataiku 4.1 now supports
reproducible environments, which properly isolate projects and reproduce
the runtime condition throughout the deployment phase. This alleviates
the worry about one individual upgrading a package, because deployed
code will remain stable.
Dataiku 4.1 bolsters the product's end-to-end reach by introducing a
versatile API node that scores models, runs custom Python and R
functions, and accesses to datasets via parameterized SQL and custom
functions. Additionally, the new release provides an extended toolkit