Modern factories are data rich environments bristling with controllers, sensors, and other computerized gear. The problem is there are too many data sources but not enough real-time flow of that data to the people who need it to ensure smooth operations.
That's the rather large problem that Alluvium wants to address by applying machine learning to that data.
Now, the thus-far stealthy New York-based startup has $2.5 million in seed funding to boost its efforts. Investors include IA Ventures, Lux Capital, Bloomberg Beta, and Mike Olson, chairman and chief strategy officer of Cloudera.
Machine learning is technology that can glean patterns from huge amounts of data and make useful predictions based on those patterns in applications as diverse as fraud detection and predictive maintenance. How useful is it that a company can replace a part on an expensive machine before that machine actually fails? Machine learning, by detecting anomalies in the performance data of that machine, can facilitate that.
Alluvium's Floodplain software claims to take data from myriad points—all those aforementioned factory floor sensors, controllers, computers—aggregate that, and provide real-time feedback.
Floodplain, which runs on-premises, initially acts as "a one-way pipe" from all those data sources, to the operator," Alluvium founder and chief executive Drew Conway tells Fortune. The goal is to make even older factories as efficient as modern entities like Amazon.com (amzn), which are able to tell you want you want to buy before you buy it because they are built upon massive, recent investments in big data systems and analytics.
Get Data Sheet, Fortune’s technology newsletter
Cloudera's Mike Olson is a true believer. The Alluvium team, he says, understands the technology and related analytics, but are also focused on packaging all of those capabilities so larger companies will buy it instead of trying to build it themselves.
For more on the Internet of things, watch:
The advantage of Floodplain over broader industrial solutions like, say General Electric's (ge) Predix, lies in its promised ability to channel data to operators in real-time, Alluvium proponents say.
Shivon Zilis, partner and founder of Bloomberg Beta, an early stage investment firm backed by Bloomberg LLC, says that the industrial Internet of things has no shortage of data, but it couldn't make it to the operators in useful or timely manner until now. The ability of machine learning, she says, to separate the key data from extraneous noise has changed that.