Altair Shows Off Converged Analytics Lineup
If you’re in the market for analytics or machine learning software, you may want to keep your eyes on Altair Engineering. Best known for its product simulation and computer aided engineering software, Altair has quietly assembled an impressive big data platform that extends from data preparation and business intelligence to streaming analytics and AutoML.
Altair Engineering traces its roots back to 1985, when CEO James Scapa, George Christ, and Mark Kistner founded the Troy, Michigan company to develop CAE software. Its initial product, called HyperWorks, allowed designers to simulate manufactured products, whether it’s a car chassis or an airplane wing.
With its proximity to Detroit, Altair enjoyed a good relationship with the automotive OEMs. The HyperWorks product was instrumental in helping the automakers simulate how their designs would behave in the real world, thanks to an array of modeling tools and digital physics engines that it had on offer.
Under the direction of CEO Scapa, Altair made a slew of acquisitions that bolstered its position in the product simulation space: solidThinking, SimLab, ACUSIM Software, and Visual Solutions, among others. The simulations typically benefit from massive computing horsepower and so the company naturally grew into the high performance computing (HPC) market. Today its PBS Works software unit, which develops software for managing HPC clusters, is present in nearly 40% of the supercomputers on the Top 500 list, the company says.
As the volumes of data generated by the simulations and other sources grew, the need for big data tools increased for Altair’s customers. And as the capabilities of machine learning grew, so too did the temptation to take advantage of ML capabilities on HPC hardware. Increasingly, enterprises wanted to pair their traditional HPC simulation solutions with newer machine learning and artificial intelligence (AI) capabilities.
Scapa detected this market shift and directed Altair into a new direction: big data analytics.
In 2018, just following a successful 2017 IPO, Altair paid $176 million to acquire Datawatch, whose flagship product, Monarch, provided data preparation and business intelligence capabilities. However, prior to its acquisition by Altair, Datawatch made a pair of its own acquisitions, including a Swedish developer of streaming analytics called Panopticon software back in 2013, and the 2018 acquisition of Angoss, which had 20 years of experience in machine learning and data science.
Today, those Datawatch products form the core of Altair’s big data analytics strategy. The company, which reported $495 million in total revenue last year, sells four main products in data analytics unit, including:
- Monarch for data capture and desktop-based data preparation;
- Panopticon for real time visualization and stream processing;
- Knowledge Hub for browser-based data prep;
- and Knowledge Studio (formerly Angoss) for machine learning model development.
Thanks to its Datawatch acquisition, Altair made it into Gartner’s most recent Magic Quadrant for Data Science and Machine Learning, although it has a way to go before it will be considered a leader by that analyst group.
But according to Srikanth “Sam” Mahalingam, Altair’s chief technical officer for HPC and cloud products, the company’s data analytics offerings are uniquely positioned thanks to the breadth and depth of other Altair products.
“We are truly seeing the convergence of simulation, HPC and analytics,” Mahalingam told Datanami in a recent interview. “We are bringing these tools together and truly providing a platform for our customers…to make effective design decisions and shrink the time to innovation,” he said.
During its virtual Global Data Analytics Summit last month, the company launched new releases of all of its data analytics products (Mahalingam noted that the launch came on the 30th anniversary of the launch of the initial Monarch product, “right from the day it was developed as a DOS-based tool,” he said).
Monarch benefits from improved handling of data sources, including CSVs and Excel files. Better handling of complicated Excel data allows users to perform data prep work in the desktop Monarch environment instead of Excel, and even duplicate that work on other Excel files thanks to an extensible template. Extraction of data from PDFs – a longtime Monarch specialty – has also been bolstered.
Knowledge Studio, the former Angoss product, has also been enhanced, with support for the Keras deep learning framework, auto-generation of machine learning models in Python, and support for explainable AI.
Keras is one of the most popular open source frameworks for developing neural networks, which is why the company added support for it in its drag-and-drop machine learning tool. While the offering itself is not open source, Knowledge Studio leverages other open source tools, like Apache Spark, and it will likely adopt more in the future, Mahalingam said.
“We are not rewriting what the open world is doing,” he said. “But whenever we feel the open world is providing the best of the tools out there, we are integrating with them, because many of these large enterprises have standardized on these open tools.”
Knowledge Studio can automate large swatches of the machine learning model, according to Mahalingam.
“We are able to train [the model], generate the code, and that particular Python code can be deployed as an inference or scoring model into any of the enterprise platforms, as well as deployed into our own operational platform in order to build meaningful applications,” he said.
The addition of explainable AI capabilities are necessary to allow citizen data scientists to trust the models that are automatically generated by Knowledge Studio, Mahalingam said.
“If we don’t provide explainable AI, then the business user is not going to have confidence to use an auto generated model,” he said. “That’s why we feel that explainable AI and AutoML go hand in hand.”
Models developed with Knowledge Studio can be deployed into Panopticon, the real-time analytics engine. This will be useful for users in the marketing and finance department, as well as those in product engineering and design, which is Altair’s traditional sweet spot.
“If you are building a real time visualization and you want to have your scoring models score the real time data that is coming in to put together a preventive maintenance application, and you want to augment with continuous intelligence, we’re able to do that as well,” Mahalingam said.
Altair may not be a household name in the data analytics space. But thanks to its heritage in HPC, substantial customer base (11,000 and counting), and proven ability to absorb acquisitions (it has conducted 30 to date), customers don’t have to feel like they’re risking their businesses on an unproven startup.
“Today we can provide a complete end to end enterprise platform to quickly build analytical applications, whether it is in the traditional business environments like marketing or finance or …in industrial IOT and product engineering,” Mahalingam said. “We can bring in a lot of experience and expertise and are able to now converge both the product engineering, the industrial IOT, and the business analytics. We are able to bring all these together with Altair under a single umbrella.”