By Kevin McCarthy // April 28, 2017
Daniel Wilson is a Postdoctoral Fellow at Lawrence Berkeley National Laboratory. He recently launched a social enterprise, Geocene, based on his research funded by the Development Impact Lab and the Center for Effective Global Action. Geocene will design and build the next generation of an innovative IoT data-logging sensor. Originally designed to monitor cookstove usage, Geocene’s sensors will be generalizable to many more tasks, allowing for measurement of metrics like pressure and vibration. We recently caught up with Danny about his new endeavor:
1. Tell us about Geocene– What is its mission, and how did it spin out of your work at Berkeley?
Geocene makes it easy for anyone to deploy wireless sensors and analyze the big data they create. Geocene’s mission is to use simple sensors and amazing analytics to enable the Internet of Any Thing. Today’s Internet of Things (IoT) companies are solving hyper context-specific problems by pairing sensors with single-purpose mobile and web applications to create niche products. This leads to narrow use cases such as a temperature sensor that controls your furnace, a wearable temperature sensor that monitors your baby’s health, or a probe thermometer that sends you a push notification when your steak is perfectly cooked (you can’t make this stuff up). However, a general-purpose temperature sensor and analytics for a generic use case does not exist. Today, if someone wants the convenience of IoT products to measure or actuate the physical world, they must hope a purpose-built hardware and software product exists to satisfy their needs (thank goodness for the Green Mountain Davy Crockett Wi-Fi Grill with Sense-Mate Thermal Sensor). If you want to use sensors for generic applications, or in markets that are not attractive to tech entrepreneurs, the promise of the IoT remains out of reach. Geocene wants to be the antithesis of the hyper context-specific sensor industry by building rugged and simple data-logging sensors that can be deployed in fleets of 1 or 10000, with the resulting data analyzed using machine learning. Because Geocene’s software system is integrated and intelligent, customers don’t need to write any code or know any data science to process large data sets.
Geocene grew out of my Ph.D. research at Berkeley on monitoring and evaluation of cookstoves. In my work, I deployed hundreds of commercial data loggers on cookstoves in Sudan and India to measure whether or not the cookstoves were being used. Throughout this process, I became keenly aware that the market did not have good solutions for rugged & low-cost data loggers, especially if those loggers needed to be deployed in large fleets. Also, I was overwhelmed at the challenge of processing the millions of data points that a large-scale sensor-based study produces. The process of deploying sensors on cookstoves and processing their data took me an entire Ph.D., but I wanted to enable other people with less time and less training to deploy sensors and process data easily. That’s why I turned my dissertation work into a social enterprise that makes the hardware and analytics I would have wanted.
2. What are the technical innovations of Geocene’s new sensor, the Geocene Dot?
Geocene is innovating in both hardware and software. First, our Geocene Dots are ultra-rugged waterproof sensors and can live for more than a year on battery power. Our sensors communicate to a mobile application over Bluetooth, letting customers easily tag sensors with notes, metadata, and GPS coordinates. This application simplifies data collection across hundreds of thousands of sensors. Although Geocene was born out of the need for better temperature loggers for cookstove monitoring, our vision is to build better data loggers for a variety of industries. We plan to enable anyone to measure and process data about temperature, pressure, vibration, light, and other metrics to answer questions relevant to their unique context.
Once you have your data back, Geocene helps you make sense of it with an online tool called Studies. Studies lets people who don’t know how to write code process huge datasets with a simple graphical tool. You highlight examples within the data that are interesting to you, and Studies will use advanced machine learning techniques to find similar events in the data and give you summaries. For example, if you were using a Geocene Dot to monitor cookstoves, you could highlight a few cooking events in the temperature data as examples of what you think “cooking” is. Studies will find all the other cooking events for you, and give you summary statistics describing how often people cook, how long they cook for, what time of day, day per week, etc.
3. Can you tell us how you became interested in developing technologies for monitoring and evaluation?
I was involved with the Darfur Stove Project since I started at Berkeley. To date, the Darfur Stove Project has disseminated about 47,000 efficiency cookstoves in Darfur, Sudan. However, we wanted to measure adoption of those stoves with sensors to get better insights about if and to what extent customers were adopting the cookstoves. Originally, I saw sensors as a means to an end; I just wanted to know if and how much the cookstoves were being used. However, as time went on, it became clear there are big holes in the sensor market, and I became very interested in the field of monitoring and evaluation as a practice. I was especially interested in how hard it was for non-technical stakeholders (who comprise most of the international development community) to deploy sensors and analyze their data. My work in Darfur taught me how important and useful sensor data was, so I wanted to build tools and technologies to make sensor-based monitoring and evaluation more available to everyone.
4. How did you get involved with social scientists, and how has collaboration with them changed your views/approach to engineering?
I came to work with social scientists during my time with the Darfur Stove Project. I met another Ph.D. student, Angeli Kirk, who was working on some of the social science questions related to the Darfur Stove Project’s program, and we began to collaborate on sensor-based monitoring and evaluation. Angeli, who was advised by CEGA affiliate Jeremy Magruder, helped me to ask the right questions of myself and of my research subjects to gain better insights about how cookstoves were used. Working with social scientists has dramatically changed my approach to engineering. When I came to Berkeley I was interested in answering technical questions with technical solutions. However, after working with CEGA and DIL, I am mostly interested in how technology can be used to answer difficult questions in social science and economics.
5. How has being involved with the CEGA and DIL ecosystems helped you move your research forward?
CEGA and DIL have been invaluable resources in my Ph.D. work. Nearly every substantive insight and accomplishment I have made at Berkeley was funded or supported by CEGA or DIL. I was especially grateful for multiple small grants I received over the years that allowed me to explore new sensor or machine learning technologies with my own funding and without being tethered to a much larger and more unruly research program.
The community around CEGA and DIL are also great for support and collaboration. I have developed great peer networks from the research symposiums and seminars the DIL puts on. Some of my DIL-funded machine learning work has even made its way to another DIL-funded group in Portland, SweetSense, who are using sensors and machine learning to predict when hand-operated water pumps will fail.