Back in Pittsburgh, I had recently visited a local power plant run by a former college alumni who was suffering from persistent downtime at the facility, costing millions. This was my first experience seeing energy infrastructure and experiencing the vast scale, old structures, and poor conditions. During that visit, I learned that someone had recently fallen and died where I stood while trying to find structural defects at elevated heights. I knew there had to be a better way.

It was clear that the existing methods of doing that job were dangerous and expensive, so I started building a robot in my dorm room that could climb walls and used ultrasonic sensors to take a sonogram of entire steel structures. The robot failed a lot, and after days spent on site in the depths of dirty boilers, armed with soldering kits and my Raspberry Pi, I eventually got it going. I found the problem area and saved the company tens of millions of dollars. I was onto something.

Robotics ‘a great way to fail’

That initial wall-climbing robot laid the foundation for Gecko Robotics, and the pursuit of diagnosing the health of the built world. Now, Gecko builds robots and software to power our AI-driven operations platform, Cantilever, which manages over 500,000 critical assets, from power plants to bridges, and even ships for the U.S. Navy. 

But it wasn’t an easy journey after that first visit to a power plant. Startups don’t typically build tech for industrial clients, and starting a hardware robotics company was “a great way to fail,” according to many VCs I talked to along the way. I went to some pretty dark places, both physically in the industries I was trying to help, and mentally too.

For years, I was bootstrapping, sleeping on a friend’s floor, and living on the edge. I poured every penny I had ever saved into the company—and I was down to my last $100. I held on way longer than any rational person would have because the mission was noble and no one else was trying. So I kept going. 

In late 2015, a friend from college who went on to Carnegie Mellon University told me to talk to a venture partner from a startup accelerator called Y Combinator who was coming to his campus. That friend was Troy Demmer, my eventual cofounder. That was a game-changer. Gecko received the standard $125,000 investment for a 7% stake, meaning the initial valuation was nearly $1.8 million.

An investment offer with a catch

I couldn’t believe it, but more challenges were on the horizon—ones that really tested how much I believed in Gecko. A few months later, a well-known billionaire came forward to make me an offer—a $2 million investment that would have valued Gecko at $10 million. I was blown away.

There was a condition, though. I had to move the company to San Francisco and build the robots in a lab there to achieve full autonomy before launching to the world. In 2016, this way of building was the prevailing wisdom for robotics: Build a fully autonomous robot in Silicon Valley, then launch. You would be a fool to do otherwise; both the talent and capital were there.

Read more from Fortune

However, I knew it was wrong for me. It went against everything I had created and wanted to build: forward deployed robotics engineering while being with and close to the customer. In the case of understanding the needs of our customers and working with them, robotics needs to be built in the field, like I did for the very first time at the power plant back in Pennsylvania, where you can test and iterate in real-world conditions.

This felt like the biggest moment of my life, and maybe the only investment of that size I would ever get. But all I thought about was how the mission would probably fail with me sitting in a nice lab in California. I knew I belonged in the dirty, dark places where the forgotten working class existed—an environment that software companies and Silicon Valley never helped.

I turned it down. It was a huge risk, but I believed in the Gecko way of building.

Fast-growing robotics startup

Gecko kept moving forward, slowly at first, but with a growing team and more customers. We started to show that focusing on the problem first, not the solution, gave unfair advantages to our customers. Mining companies, steel companies, power companies, and government (for their defense assets and bridges) began working with us—some of the biggest names in the world.

Today, Gecko has over 300 employees in five major cities in the U.S., as well as the Middle East, and we are doubling in size this year. We’re doing it with our headquarters in Pittsburgh, not San Francisco. But more importantly, we have the largest repository of data held on the built world and are working with the biggest companies to understand even more.

My journey so far has been a rollercoaster, full of setbacks and near-death experiences (both physically and company-wise), whether that’s escaping from falling scaffolding while visiting a power plant in Hong Kong, or when a screwdriver smashed in front of my feet after it dropped from a worker 150 feet above me in Kentucky. 

But it has also been incredibly rewarding, as I learned that grit, a clear vision, and a belief in something bigger than yourself are essential for building something truly impactful.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

Read more:

• Beyond the humanoid hype: Why specialized robots will dominate the next decade
• How we built our bootstrapped startup different and sold it for $40M. (Hint: We ignored some myths)
• At 18 we cold-pitched Etihad on upending air miles. It worked, and 3 years later we’ve raised $1.5M to transform travel loyalty programs
• How our ping pong startup hit a $50M valuation in 5 years by tapping into automation