Fraser: Jason, talk me through your early years. How did you get into software?
Jason: I took an odd path. I have an East Asian Studies degree and most of a literature major before I dropped it. So I actually got into software in 1997, having learned some German and some Japanese, and using what I had learned about the internet at that point to land a position at Microsoft.
My unusual background in language somehow allowed me to start doing internationalization testing for Internet Explorer 4 before becoming a software design engineer. After which I went in the direction of building complex software solutions at the Seattle Times and other companies as a consultant. Eventually, I moved to Japan in 2017 primarily so I could take care of my father-in-law here and my kids could attend elementary school.
Fraser: Having established a base in Japan, how did you meet Arjun Chandar and end up working together?
Jason: I spent a lot of my time initially in Japan networking and meeting people. At one of these tech network events, I met someone who had an exciting company proposal and was planning to move to Japan. When they asked me about the prospect of some part-time consulting, I was intrigued. And, of course, this person was Arjun.
He wanted to know if I knew anything about extracting data from factory devices, and my response was quite plainly, “not really, but it sounds right up my alley!” After a few months of negotiation, where we were waiting for Arjun’s development partner, Daiwa, to finalize a contract, I signed up as a consultant and began building all the initial prototypes and proofs of concept.
Fraser: What aspect of the proposal appealed to you?
Jason: When Arjun proposed the broad concepts of Machine Learning (ML), power management, maintenance, and defect detection, I couldn’t help but start formulating ideas on optimizing power utilization and help IndistrialML make the right purchasing decisions for electrical equipment that needed replacing.
And I knew it was time for me to put all of the experience I had accumulated in software over 20 years into practice. I could also really challenge myself by asking: “Can I start a whole system from scratch and grow it organically and integrate new team members into it?”
Fraser: What did IndustrialML and Daiwa want to create, and how was it developed?
Jason: During my first few meetings with Daiwa on-site, I quickly realized that data was not being properly harnessed. I remember asking about how and where data is currently being collected. The response was: “Well, we have a programming logic controller (PLC) that stores some measurements on an SD card, and at the end of the day, we may go and pull it into Excel and take a look at it…”
So I immediately thought it would surely be helpful for factory floor workers to translate that data into visualizations in a format that can show you historical trends over time. Then I began building a prototype of an ingestion system for capturing that data. The process was a collaborative back-and-forth effort to get some data into a system where we could show simple charts and visualizations through dashboards.
Fraser: What were some difficulties you encountered?
Jason: To get to the point of creating a finished product, we had to add a more sophisticated alerting system that would support multiple destinations and increasingly complex rules. This has meant almost four major rounds of changes.
Initially, we used a GoPro to test the value of capturing video on the line from a video monitoring system. Meanwhile, I was writing code that supported capturing data from an industrial camera and aggregating the buffers into a video stream and pipeline, which was a completely novel concept to me. Finally, there was also the small matter of integrating the product with a third-party audio service, allowing us to correlate events happening on the line with discussions between factory workers.
Fraser: What do you see for the future of growing the product?
Jason: As we correlate a mix of time series, enterprise resource planning (ERP) data, and other third-party systems, we’ll come to a point where growing into a full suite of tools that can scale with disparate data types will be a serious challenge. I think that our ability to solve those problems will be vital in the evolution of the product.
Fraser: Thanks so much for your time Jason.
Jason: No worries, it was a pleasure chatting.