We knew AI was hot, but the last week has brought several announcements that show the industry is trying to bring AI and machine learning (ML) into the enterprise mainstream. With two AI acquisitions by big names in the analytics space and a $100M funding round for an MLOps specialist, this has been a week of AI acceleration.
Today's news is that Databricks is acquiring German company 8080 Labs, makers of bamboolib, a UI-based data science tool that generates code, but doesn't require users to write any. And this follows the announcement earlier this year from Databricks that it had added AutoML capabilities to its data "lakehouse" platform.
Also read: Databricks ups AI ante with new AutoML engine and feature store
Clemens Mewald, Databricks' Director of Product Management, Data Science and Machine Learning, spoke to ZDNet and explained that bamoboolib is presently offered as a Jupyter notebook plug-in and will, logically, be integrated with Databricks' own notebooks. Afterwards, it will also be integrated into the Databricks workspace user interface, to make it available to less technical users, often referred to as "citizen data scientists."
Mewald explained that the acquisition aligns with Databricks' approach to expose multiple layers of abstraction: at the coding level for data scientists, machine learning engineers and, of course, data engineers; at the SQL level for more database-oriented professionals as well as business power users; and at the UI level for those users less technical yet still passionate about their data and analyzing it.
While we're on the subject of integrating AI into data analytics platforms, that's what Qlik had in mind when it announced its acquisition of Big Squid on September 30th. Big Squid offers an AutoML platform which will continue to be be offered standalone, as Qlik AutoML, but will eventually be integrated into Qlik Sense. This will allow BI (business intelligence) users to use their data sets for training models, then scoring additional data against the models, with predictions brought back as new columns in the data set, where they can be easily visualized just like any other data.
Big Squid's technology even provides ML model explainability, via use of Shapley values, something Databricks' AutoML does as well. The fact that both vendors provide Shapely value-based AI explainability is another sign of AI's enterprise ambitions, since certain data protection regulations require this of AI models, in effect as a matter of trust and auditability.
Qlik's goal, much like Databricks' is to make AI accessible to analytics teams and not just data science teams. More specifically, Qlik is focusing on key driver analysis, predictive analytics and "what if" decision planning, all integrated into the BI environment and paradigm. And, yes, Qlik also wants those coveted "citizen data scientists." Plus, armed with the technology derived from its many other acquisitions, like Podium Data (now Qlik Catalog) and Attunity (now Qlik Data Integration), Qlik's presence across the full data lifecycle, and its governance, gives the company a good MLOps (machine learning operations) play too.
Speaking of MLOps, that's the very area focused on by Domino Data Lab, and specifically in the enterprise context. And, just yesterday, Domino announced its $100M Series F funding round, led by Great Hill Partners, with existing investors Coatue Management, Highland Capital Partners and Sequoia Capital and new investor Nvidia. Nvidia, of course, is the premier vendor of GPU chips and servers, all of which can accelerate AI training and scoring, especially in the realm of deep learning.
In fact, Domino and Nvidia used yesterday's funding announcement to reveal their own enhanced partnership. According to its press release, the partnership will involve Domino's development of product functionality to expand the accelerated computing capabilities in its platform, including "validating the Domino platform for NVIDIA AI Enterprise so that Domino can run seamlessly on mainstream, NVIDIA-Certified Systems from OEM hardware providers." Domino also says the two companies will bring these products to market together.
That all three of these deals were announced in the space of a week shows how important AI/ML is becoming in the enterprise computing context. It's possible -- even likely -- that this first wave of AutoML integration into BI and analytics platforms is just an early iteration of the AI + BI equation. Automation of ML algorithm selection, hyperparameter optimization, feature selection and even explainability is great. But eventually there will need to be automation of feature engineering, model deployment, monitoring and retraining too.
Also read: DotData boasts automated feature engineering for Databricks
Things are happening quickly enough that these first-gen innovations may become proverbial "table stakes" quite soon. As analytics companies then move into competitive differentiation, the AI world may become more interesting, relevant, accessible and, hopefully, more easily scrutinized and trustworthy, too.