TellusLabs Provides a Living Map of the World’s Food Supply
Do you ever think about where your food groceries come from? How about home cleaner, deodorant, and face cleanser? Did you know that all these products share the same core ingredients and that these ingredients come out of the ground? The whole consumer products economy relies on a complex supply chain that starts with agriculture.
Somerville-based TellusLabs makes technology to map the world’s food supply. Founded in 2016 by David Potere (also CEO) and Mark Friedl (also CSO), the company utilizes machine learning and satellite data to build a decades-long portrait of the Earth’s surface. Their first product, Kernel, is an agriculture-focused platform that provides, “a suite of agricultural intelligence insights related to grains and oilseeds around the world.”
I had the chance to sit down with Potere to learn more about the company’s story, as well as the details behind their AgTech solution.
To start, tell me about your background.
I fit into the TellusLabs story primarily through the technology. I started learning about mapping technology and satellites in the Navy—I was a minehunter from 1998 to 2002. I came out of those four years of service just totally enamored with geospatial and satellite tech, and I really wanted to find out what was going on under the hood.
I really fell in love with the power of remote sensing tech, so I signed up for what started out as a master's program at Boston University in 2003. The very first Professor I met on my very first day in the remote sensing program was Mark Friedl, who is my co-founder. So Mark—our chief science officer—he’s been stuck with me for like 15 years now one way or another.
He’d be the last guy to say it, but he really is a pioneer in the field of satellite remote sensing. When I met him back in 2003, he was building the very first operational map of the whole planet. He would take satellite data every day, and on a monthly basis, build a fresh map of the planet. I met him in the middle of doing that, and it is now the backbone of how we visualize the surface of the Earth for climate models.
So long story short, I finished my PhD in Princeton, I joined Boston Consulting Group, and I built out the the geospatial team which eventually became their data analytics function, Gamma. A lot of the early joiners that came with me to build out that sort of geospatial intelligence unit inside BCG I made come over to TellusLabs. And Mark and I built the TellusLabs team with the talent networks we built over our careers.
What brought you to the aha moment behind TellusLabs?
There was no shortage of customer problems where I knew imagery could matter, but our technology was not well suited for the typical engagements at BCG.
And for my co-founder Mark, he was leading multimillion-dollar NASA projects but saw a gap where a larger impact could be made. He reached out in 2015 to me with a prototype that is predicting crop yield in the US better than the USDA. The two of us then worked every weekend for the better part of a year on a prototype model to show the power of operational daily scale satellite models, and our target was US corn production. We decided to enter MassChallenge, and I finally left BCG in March 2016 once we realized that our models were beating the US government’s own forecasts pretty regularly.
That summer, we started daily forecasting for corn and soy, and it was like a daily email for Mark and myself. At first, it was for our friends and family every morning at 9 a.m., but by the time the summer was over, we had a thousand people following us every day. So that's what we took to the venture world, we closed our seed round in the final days of 2016, and our first hire was in January 2017.
How do you get this data?
Most of the raw data we use for TellusLabs can be downloaded freely. We use what's known as open data, which has been generated by NASA and the European version of NASA called the ESA. They’ve been building these kinds of images for 23 years now, and the archives have been available for that whole time.
What's challenging is translating the raw imagery into a useful signal, so a lot of the secret of our success at TellusLabs is the combined 50 years of experience for designing, building, operating, and extracting value from the satellites. Any given moment, 70% of the earth is covered by clouds. The atmosphere is thick and creates a lot of noise. The satellites look at the ground from a wide variety of angles and one satellite is almost never enough, so we draw from a half-dozen satellite systems now to build a single view of the ground. A lot of the art here is combining those different signal feeds.
Who are the primary buyers of this data?
So what Fernando [Rodriguez-Villa]—our head of commercial at TellusLabs—would tell you is that the initial wave of customers that we found in the first year were heavily concentrated in the financial sector—hedge funds and asset managers. And we're still proud to have a whole customer segment around that space.
What’s happened as the company has matured is that we’ve found a whole universe of other brilliant, important Fortune 500 companies that depend existentially on food. These include both agriculture-oriented businesses and consumer-packaged good (“CPG”) companies, so these companies depend existentially on the corn and soy harvest. They have whole teams dedicated to helping them source the grains they need efficiently so they're not getting hurt by market fluctuations or by unknowingly having unsustainable practices. The transparency our tech is providing will allow for future fertility in the land that will enable our clients to continue to deliver value.
Do they buy this on a subscription?
Some subscribe to our Kernel SaaS product, which gives them access to both a web app and API, which allows them to programmatically pull the data they need. Increasingly we’re seeing demand from enterprises to integrate with the underlying platform itself. This is really exciting for companies that have their own data or set or use cases. These integrations take some configuration work up front, but they result in really special client experiences of the technology.
Did technology have to catch up with the ability to process this information?
Yes, several technologies needed to come of age for TellusLabs to be born. For us, cloud computing is essential. A lot of this data was stranded because it was just at a scale that you couldn't access fast enough, and in fact, Amazon and Google have worked to make satellite imagery, in particular, more available with lower latency than was ever possible before.
The second necessary thing that would have made this impossible five or six years ago is that machine learning had to mature. When Mark was starting this adventure back in 2003, he was having to hard-code most of the algorithms that he used for the model builds.
The last thing is that we needed a long enough historical dataset for our models to learn from. Although we have been watching for 30 years, it's only in 2003 that we started watching everywhere, every day, four times a day. We needed to do that for about a decade before we had enough training data in order to do something valuable in agriculture because you need a lot of reps, and a decade's worth of data turns out to be about the minimum you need before you can make trending models.
Your company has a very compelling social mission attached to it. Can you share why it is important to have a digital map of the world’s food supply?
For one, it's because that map is changing fast, and it’s absolutely critical that we have technology that keeps up with it. Whether we acknowledge it or not, changes in our environment and demographics ARE rewiring where food gets grown, so EVERYONE will need a digital map like TellusLabs’ to be able to react to them, plan for them, and to make sure we keep providing food.
Another trend is that consumers are demanding traceability. They want to know where their food is coming from, and really, part of the chain of transparency needs to start by looking at the field images that are consistent and reliable on where the food was made.
Lastly, farming is right on the edge of being economically sustainable. The whole sector is changing right now because of shifting information advantages between farmers, major corporations, and consumers. It's crucial that there's an objective game board that's available in order to keep the economy sustainable.