Krishna K. Gupta founded Romulus Capital in his MIT dorm room when he was just 20 years old. In the last nine years, Gupta has built Romulus into an active, early-stage venture capital firm that invests primarily in technology and science-enabled companies, including Cogito, Humanyze, Ceres Imaging, and Allurion Technologies. With approximately $200 million in assets under management, the firm has backed more than 30 companies.

“I went to MIT and fell in love with the idea of helping my peers build businesses,” Gupta says. “This was back in a time when venture capital activity was far less. Seed stage wasn’t really its own asset class. Most importantly, I felt that most investors were betting on companies instead of helping build them, so that was the gap I wanted to help fill.”

In a conversation with Fortune, Gupta talks about the evolution of the venture ecosystem over the last decade and his firm’s contrarian approach when it comes to investing in early-stage companies.

FORTUNE: You launched Romulus Capital in 2008 at the height of the recession. From your perspective, how has the venture ecosystem shifted since then?

GUPTA: It was not much of an ecosystem then. There was very little activity. One of our earliest companies came out of the MIT Media Lab. We were the first $40,000 check in that company, and we couldn’t find anyone else to put in another $100,000. Today, the same company with the same team would’ve probably raised $4 million in a week from some West Coast VC.

The only people starting companies in that kind of bust period are really serious about it. So the talent becomes more clustered. For instance, if you have four people who are really good at what they do in a 2008-type period, they come together to start one or two companies. Today, they’d each start their own companies. So you now have much more fragmentation of talent, you have more competition, and you have much higher valuations. That’s a triple whammy that doesn’t really help you out as an investor. I actually really liked 2008 — I felt that the conversations I had and the types of entrepreneurs I met were just much more audacious and authentic.

What types of companies did you invest in back then?

GUPTA: We were very interested in where we could take the intersection of machine learning and human behavior, which over the years evolved into emotional AI. Our first investment in machine learning was in 2010, which put us ahead of the curve in that space.

We don’t like to do momentum investing. If you’re looking to build companies in the long run, you really need to have a contrarian perspective. If the venture scene zigs, we zag.

Why do you think the approach of older, more traditional VC firms is problematic?

GUPTA: Two things are happening. One is that these venture firms have gotten so large that they’ve become more like asset managers. Their business has become more to deploy capital at a relatively fast pace. That’s not right or wrong, but it’s just a different business than what I perceive to be venture capital.

And two, there’s so much competition that people are thinking, “If we don’t invest right now, we’ll miss out on the opportunity.” Everyone is pressured to do deals, but the reality is there aren’t that many great ones. Yet people are still deploying capital. So what you see is the weird effect of herd mentality where everyone runs toward the same thing. The AI wave is starting to subside now, but for a good year, it was all AI, AI, AI. Whenever I hear everyone around me doing the same thing or saying the same thing, I tend to want to run away from it.

Speaking of AI, there are a number of companies working on building a brain-computer interface, which would allow the mind to connect with artificial intelligence. What do you think about the future of these innovations?

GUPTA: Broadly, we understand very little about the brain. The intersection between machine learning, neuroscience, and behavioral science is really potent. There’s going to be some really great work to come out of that. Companies like Facebook, Neuralink, and Kernel are trying to figure out how to map and sensor up the brain, but I think it’s relatively long-term before we can see truly impactful work. If you take the neuroscience out of the picture, and you focus on machine learning at the intersection of behavioral science and psychology — now that is where I think there will be a lot of powerful work going on.

For context, we’ve invested in two companies called Humanyze and Cogito. Humanyze helps companies better understand how their people are communicating and interacting, and Cogito uses emotional AI for voice to be able to help call center agents react to customers in real-time. In those kinds of use cases, I think we’re unlocking the power of what it means to be human.

What are some trends you’re seeing in the venture community as a whole?

GUPTA: I’m not the first one to say is, but we’re living in an insatiable time. You look at what Softbank is doing; pumping a lot of money in the ecosystem. The most important thing for founders to know is that this won’t last forever. People need to understand how to build fundamentally strong cultures. In these kinds of environments, that gets lost.

In 2008, you might’ve gone out to raise $2 million, and it turned out you could only raise $500,000. And you know what? You make do with that. Today, you go out to raise $2 million, and you end up with $5 million. And you say, “Oh well, OK, I’ll take the extra $3 million.” That almost always destroys your culture from Day One, because you’re not optimally spending your resources. If that’s your DNA from Day One, then even when you have $100 million in the bank, you’re going to sub-optimally spend your resources.