This Data Scientist Is Using AI to Revolutionize the Farming Industry

If you’ve ever felt discouraged because you’re not quite sure what type of career you want, take it from Sam Swift: You will figure it out. And he should know. Swift worked in software and academia before taking data science roles in national intelligence, fintech, and, now, agriculture.

“I think it’s OK not to have a very specific plan for your career because it’s quite likely that you haven’t even heard of your dream job yet,” he says. “That’s the best way I can explain my path.”

His latest role is as Senior Director of Data & AI at Bowery Farming, which grows vertical, sustainable indoor farms outside of urban areas. Swift oversees the analytics, machine learning, and software teams. “I maintain a broad perspective of how technology can enable our farms to thrive, and work to connect the technology vision with the business needs and operational realities to solve challenges,” he says.

Here, Swift talks about how Bowery Farming is revolutionizing its industry, what he loves about the company culture, and how to succeed at a startup.

I graduated from Carnegie Mellon University as part of the first class of a new major called Decision Science. I then worked at a small software consultancy as a developer, learning software engineering on the job. After a few years, I couldn’t resist the intellectual challenge of academia and returned to CMU for a PhD in organizational behavior. I studied the intersection of economics and psychology applied to managerial decision making and negotiation.

From there, I happened upon a number of roles doing data science before I’d heard it called that. I took a job leading the data science team at an academic spin-off startup called the Good Judgment Project, which was competing against the CIA to predict geopolitical events using wisdom-of-crowds techniques and machine learning. I realized that I enjoyed building the system and leading the team more than writing the papers.

I started looking for true industry jobs, and found a great fit in New York City with Betterment as their first data scientist. I had already been an enthusiastic customer, and the personal finance domain meant my academic background had crossover value. After three-plus great years, I was eager to test my experience against a novel problem space. That’s when I heard about Bowery.

Bowery is a unique opportunity for a technologist with broad interests. I was intrigued by Bowery’s thesis being really centered on software and data opportunities, but working in a fundamentally physical business. I was craving more exposure to the “real” world because it seems like a new challenge, but also because I think it’s a blind spot for much of the tech community. You can understand a lot of how the world works these days through the lens of purely digital businesses, but the physical world isn’t going anywhere. I wanted to try my hand at crossing the divide and learn from a new community of experts in agriculture, operations, and engineering.

Of course, connecting with the physical world was not the only draw. If it had been, I could have gone into the oil industry or consumer goods. The other huge factor is that Bowery is a company I’d want to be successful even if I didn’t work here. I don’t have to twist myself into knots to explain why spending my time and energy on this idea is worthwhile. It’s very motivating to bring home our product, put it on the dinner table, and know that if we achieve our scale aspirations we’ll be doing so much for the sustainably produced fresh food supply.

Over the last three years at Bowery we’ve built an increasingly complete and complex system to operate our farms while scaling larger and larger. The technology that powered our first farm was very impressive, but also somewhat ornamental. The scale of that first farm meant that it was still possible to manage by walking around, checking on things, and mentally keeping track of farm and crop state. Our subsequent farms in New Jersey and Maryland are orders of magnitude larger, which changed the game. With their scale, the technology was absolutely critical to be able to operate every day.

We have succeeded in leveraging technology to operate these vertical farms of unprecedented scale, but that means we’re fully into the next chapter. Now it is up to us to demonstrate the full potential of agriculture built this way. Every month we’re growing more types of crops with more customization and further optimization to maximize productivity and quality. This is when it really gets exciting.

When developing solutions in a very novel space, it can be difficult to judge how big each new problem is. I’ve underestimated the complexity of new areas of work on a couple of occasions and attempted to bite off way too much on the first pass. This can make very promising projects feel like busts or mistakes, and can be demoralizing for those who contributed. It can also blow up timelines we’d promised to stakeholders. But converting these kinds of mistakes into progress requires only that we learn more about the problem in the process. Our business is tackling interesting problems that don’t have well-defined solutions, so one of the most valuable things we can achieve is clarity about the problem structure. Whenever we realize we are in too deep, the goal becomes to clearly define the smaller building block projects we can chunk out to make progress against the big picture.

Bowery is the most interdisciplinary place I’ve ever worked, and it defines the culture. Few problems can really be solved by a single functional area, so meetings with agricultural scientists, operations leaders, and roboticists all at the same table are routine. Of course, that diversity of professional expertise means that the team reflects a wide variety of training and work cultures that don’t automatically mesh. I’ve seen that the people who thrive most at Bowery find it energizing to learn new lingo, respect unfamiliar customs, and reach understanding before rushing to push a view.

I’m not sure I know too many people who have any conception at all about what it’s like to work in agriculture from a laptop in Manhattan. In many cases I suppose the reality is less all-or-nothing than people might suspect. The technology team does go to the farms and don food-safety protective equipment as often as we can, but usually we’re not “in the fields” doing our work. We get free produce pretty regularly when we’re in the office and there is surplus, but it’s not a lifetime supply of all-you-can-eat arugula. And while we do have an amazing amount of measurement and control over every one of our thousands of growing crops, plants are complicated organisms and we can’t just make them do anything we want.

The startup world puts a real premium on near-term progress and subjects teams to constant change and new concerns. These forces can make every moment seem like the end of the world, and there is often the temptation to change direction, sprint all-out, or even to jump ship. The quote, “Most people overestimate what they can do in a day, and underestimate what they can do in a month”—or perhaps more simply, “it’s a marathon not a sprint”—speaks to me about finding success in these environments. Show up every day, make all the best decisions you can, keep the big-picture goal in mind, watch some of the distractions sail by, and when you’re tired go to sleep. You can get pretty far that way.

I would recommend taking the opportunities that are honestly interesting to you and will add something new to your skills or experiences. At every single decision point in my career, I’ve chosen the lower-paying job opportunity because I wanted it more. That’s not a decision rule I set out to adhere to, but in each situation, one of the options spoke to me and I’m not very good at saying no to that.

The good news is that following my interests has put me in situations where I can’t help but work hard and figure things out, and that’s what companies are hoping for. Good fits lead to more opportunities for growth and “making your own luck” as your career develops.