I recently caught up with Drew Conway, founder and CEO of Alluvium, to talk about the new Mesh Intelligence his new startup is creating for Industrial IoT settings. Conway, who earned his PhD at NYU, is a leading expert in the application of computational methods to social and behavioral problems at large-scale. This week Alluvium announced $2.5 million in seed funding from investors including IA Ventures, Lux Capital, Bloomberg Beta, and Cloudera co-founder, Mike Olson. Conway started his career in counter-terrorism as a computational social scientist in the U.S. intelligence community and is known for his venn diagram definition of data science as well as applying data science to study human decision making.
insideBIGDATA: Industrial IoT is a new direction for you, with your background in U.S. intelligence and later trying to apply big data to solve larger social problems. Start by telling us about Alluvium and the types of data science problems you’re setting out to conquer for industry.
Drew Conway: Industrial IoT software should be focused on providing tools and services that support decision makers in complex operations. As such, the founding of Alluvium is more of a continuation of the the work that I have done throughout my career. Companies across every major industry — from verticals like oil and gas, to horizontal business functions with lots of complexity like fleet management — have increased the amount of available data pouring out of their operations, and this trend is set to rapidly increase going forward. But no one has really focused on how to get the intelligence out of these systems, and provide high-value insights to operators in real-time.
We’re not just talking about sensors at the edge, we’re talking about the core high-capital assets OEM’s produce that are central to every industry with complex operations. We’re talking energy, manufacturing, even healthcare — the industries that affect the daily lives of millions of people — and the complex machinery that human operators have to reason with in these settings. Alluvium’s core technology provides a real-time understanding of how these operations are running, from the most local view, like a specific asset or sensor, up to the full global view, which could be a complex business operation or a whole business.
You hear a lot of talk about automation in industrial settings, and taking humans out of the loop. We believe focusing on the “low hanging fruit” of automation is the wrong approach for Industrial IoT.
The value we want to provide to our customers is to minimize risk and maximize efficiency by perfected production. To do this, we focus on making the human operators 10x more effective by providing them with the smartest tools and best information. We’re not about replacement of people, but about assisting, enabling and enhancing them to do better jobs. Most industrial operations are sufficiently complicated and human-centric that in order to be successful you still need human oversight, with the right context. We’re about increasing the intelligence of how humans reason with these systems.
insideBIGDATA: So tell us a little bit about Alluvium’s approach and how you’re going to do that.
Drew Conway: It’s a new category that we call Mesh Intelligence. We’re creating new data streams from machine data — as well as the tacit and institutional knowledge of the human operators — and bringing it into a graph structure that connects it semantically to other data assets. This is data that is typically trapped in proprietary OEM hardware.
We’re focused on unlocking the data from these proprietary systems for participation in the greater ‘graph’ of operational intelligence behind these companies. And we’re also working to bring the tacit knowledge of human operators into this graph, as new data streams.
insideBIGDATA: What does this tacit knowledge look like as a datastream?
Drew Conway: Folks who work in heavy industries, such as energy or manufacturing, accumulate deep knowledge of how to perform their work. Speak to a mechanical engineer working in a oil refinery and they can explain the intricate interdependencies of every inch of metal in that facility. Unfortunately, most or all of that knowledge stay trapped in that engineer’s head. Industries become this kingdom of kludges, where knowledge and best practices take on a tribal quality. Every team and organization stitches together its own workflow which can include paper and pencil and can certainly include a lot of tacit knowledge trapped inside an individual expert’s brain. So you get these inefficient workflows that get the job done but rely heavily on a specific person being in the right place, at the right time and interpreting complex things. We’ve seen this across every industry we’ve spoken to. If we can unlock the tacit knowledge in the expert’s head and expose it as a datastream, we believe there are worlds of opportunities for machine learning to bring new efficiencies to the equation.
insideBIGDATA: What’s the great challenge to Alluvium as you build out this technology?
Drew Conway: For the technology, it is really the nature of deployment to these industrial IoT settings. When you’re talking about edge systems, you can’t make any assumptions about the CPU, the available memory, or even internet connectivity. A new, lightweight model that brings the machine learning to the edge, closer to the knowledge workers is needed. The days of sending this data back to the datacenter for analytics to be performed and then delivered to the human operator–we feel–are dying.
The deeper, and – quite frankly – more meaningful challenge is delivering tools and services that support operators in ways that make their work safer and more productive. In industries where software tools are often some of the least valuable in a person’s toolbox, our great challenge is building systems and applications that earn the trust and love of our users by adding value to their workflow everyday.
Speak Your Mind