In the recent Duke Summer Blockchain Innovation Program, Andrew Epprecht ’22 partnered with Jack Rosenthal ’22 to found Full Moon AI, which accesses and aggregates performance data related to machines, equipment, or systems—with a current focus on elevators. Once data is stored on the blockchain, Full Moon creates algorithms to predict when components will malfunction, along with to help reduce energy consumption.

Full Moon AI won the grand prize at the program’s hackathon, earning Epprecht and Rosenthal a spot in the accelerator along existing Duke blockchain projects in more advanced stages.

“We’re now working on building our MVP, which we’ll use to pilot our solution,” Epprecht said. “Our next milestone is developing a tangible document that lays out the technical framework for our implementation in excruciating detail.”

Here, Epprecht shares his experiences and takeaways from Full Moon AI so far.

How did you hear about and get involved in this program? 

It was the FOMO that did it for me. A few of my more technically gifted friends would casually toss the term “blockchain” around in conversation, and it was enough to pique my interest, but I knew next to nothing about the technology. So I did what any curious person does: I Googled it. It wasn’t long before I became irrationally and irrevocably obsessed with all things blockchain, and my friends took notice. I read news articles, white papers, blog posts, and opinion pieces. I was listening to podcasts and watching YouTube videos hoping build a schema of knowledge that I could incrementally build upon. I was hooked. At the start of the summer, a friend of mine forwarded me a screenshot of a Duke I&E email about the blockchain bootcamp, and I jumped at the opportunity. The bootcamp was exactly what I needed to take my curiosity, excitement, and passion to the next level.

[Editor’s note: Sign up for our newsletter so you don’t have to rely on your friends’ screenshots!]

Tell me about your background. What was your level of familiarity with blockchain coming into the program? 

I’ve never personally coded a website before, but I’ve built many, including the site for one of the largest private medical practices in the nation. While my background is non-technical, I thrive in technical environments because one of my greatest skills is recruiting and leading people who are smarter than me. I learned these skills largely through trial and error, having founded seven companies, three of which failed. I was already obsessed with blockchain technology coming into the bootcamp, but I was nowhere near ready to lead a team of superstars to build the next big thing.

What challenges or problems are you most interested in? How does blockchain potentially intersect with these? 

In fragmented industries, data is heavily siloed. Large players have lots of data, so they’re able to leverage machine learning and AI to optimize their businesses. Smaller players, however, are unable to use this powerful technology since they don’t have that same access to scale. Blockchain allows us to break down these silos by economically incentivizing data sharing between network participants. Blockchain uniquely enables us to compensate data suppliers for using their data to train machine learning and AI models without letting anyone ever see the real data. Ultimately, Full Moon AI aims to create a decentralized data sharing ecosystem for AI that will someday power the future’s smart cities. Right now, the most addressable need for AI is in Predictive Failure Analysis (PFA). PFA is used to predict breakdowns of machines, equipment, or systems. Full Moon AI is working to provide early warning alerts about impending equipment failures in a variety of industries and sectors, but we’re starting with elevators. Elevator maintenance companies can use Full Moon AI to predict failures of elevators, allowing them to schedule preventative maintenance before an outage occurs, which is critical for businesses and residents.

What were your favorite aspects of this program? 

For me, the most impactful aspect of the blockchain bootcamp and hackathon was people’s willingness to help in any way they could. Outside of the scheduled Zoom meetings, I was able to connect with several Fuqua graduates who shed some light on the problems we were addressing with Full Moon AI. In particular, Alex Christian at Data Mynt helped us begin to develop a roadmap towards full decentralization of AI development on the Full Moon AI network. What this means is that in the future, we plan to allow independent developers to find their own relationships within the data and publish their own AI algorithms to the network, which they would make money from each time a business uses that algorithm. Decentralized developer access would amplify the impact of our platform since the breadth of AI models will be substantially greater than if we chose to develop all algorithms in-house. 

What are your greatest takeaways from the program, and/or how will you extend your work? 

Our greatest takeaway was the importance of conducting customer interviews concurrently with all other operations. Thanks to these interviews, we’ve already formed some fantastic relationships with several stakeholders in our early adopter audience. A few days ago, we spoke with the Director of Technology at a multi-state elevator service provider who perfectly articulated why our solution is necessary: “For most of our contracts we show up once a month, we provide service, which takes 45 minutes to an hour, and then we don’t see the elevator or know what’s happening unless we get a call because there’s a problem. So the idea with AI is having it connected so we know its performance and see where anomalies are occurring.” These relationships will be crucial in further validating our concept, getting real-time feedback, and ultimately piloting our solution in the elevator service industry.