From Microscopes to AI: How Nyad AI is Automating Biology to Protect Our Water

There’s a hidden army of people protecting our water supply every single day, and most of us have no idea they exist. They’re wastewater operators—the last line of defense before toxic wastewater flows into our rivers and communities. For decades, these operators have been doing something that sounds almost medieval: looking under microscopes to identify organisms in wastewater samples, using methods proven effective back in the 1970s.

This is where Virginia Szepietowski saw an opportunity. Through a winding path that took her from law school to competitive bodybuilding to entrepreneurship, she discovered that wastewater treatment was screaming for automation. With her now co-founder (and husband) Chris, she launched Nyad AI—a platform that uses computer vision and AI to do what operators have been doing manually for fifty years.

The result? A startup that’s not just building software; it’s transforming critical infrastructure.

The Problem Nobody Talks About

Wastewater treatment is essentially, as Virginia puts it, “a little zoo of good bugs that eat the bad bugs.” The challenge is that managing this biological process requires constant, rigorous monitoring under a microscope—and it’s incredibly expertise-intensive. Not every operator has the same level of skill or experience. Not every plant has the resources to do this perfectly every time.

The stakes are high. Inefficient treatment means contaminated water entering our rivers. Non-compliance means regulatory fines and operational shutdowns. Yet the tools available to operators haven’t fundamentally changed in fifty years.

When Virginia started exploring this world, she realized the technology to solve this problem had finally become available and affordable. Computer vision has advanced dramatically. Machine learning models are more accessible than ever. The timing was right—but building the MVP still required solving some genuinely hard technical challenges.

The MVP That Got Bought in a Conference Room

Here’s where Virginia’s story gets interesting. She and Chris didn’t start by building a perfect AI system. They started stupidly simple.

“Initially,” Virginia explains, “it was literally a rudimentary dashboard with a human in the loop being the AI expertise.”

This is a masterclass in lean startup thinking. They validated the core problem first—that customers actually wanted this solution—before investing months into building sophisticated machine learning models. Chris, the brilliant engineer, used rapid prototyping tools (Cursor, Claude, Anthropic) to iterate quickly. This speed had a specific purpose: it freed up Virginia to spend time with actual customers.

The payoff came at an industry event. Virginia presented the prototype at a lunch-and-learn session, and a customer literally tried to buy it on the spot. That’s when she knew they’d crossed the MVP threshold.

But here’s the thing that really shaped their trajectory: they learned something crucial about what customers actually want.

The Decision That Changed Everything

Virginia had an assumption that turned out to be wrong: the more features they offered, the happier customers would be.

They tried adding a third-party sensor to their solution. It would provide additional data. It would increase value. They even offered to cover the costs and handle installation. But the deal kept stalling.

Finally, in a face-to-face conversation, the customer revealed the real problem: “That sensor sounds great, but we’d have to shut down our operations while someone installs it. We’d lose production time, and we can’t justify that right now.”

Virginia’s response? “What if we just ditch the sensor and keep the software?”

The customer signed that day.

This taught her a fundamental lesson about early-stage products: simplicity wins. When you’re a startup, don’t ask customers to do too much. A single tool that does one job really well is easier to digest, easier to explain to supervisors, and easier to put on a purchase order. It’s less friction. It’s more likely to close.

Building the Roadmap Through Customer Obsession

After the MVP validation, Virginia faced the classic founder challenge: what comes next? How do you decide what to build?

Her answer is refreshingly customer-obsessed. Nyad AI set up a feedback system where customer input triggers an immediate response. There’s literally a big red light in their office that goes off when feedback comes in, and the team drops what they’re doing to call the customer.

“I am not my customer,” Virginia says. “I’m not a wastewater plant operator. The single most important thing was not building what I think they might want—it was building what they actually want.”

This is why they chose to start in Alabama rather than the UK. The US had the infrastructure and proximity to wastewater treatment plants. Virginia could spend time physically present with customers, watching them use the product, seeing where they hesitated, understanding their workflow.

As dashboards and interfaces have become cheap and easy to build, many startups fall into the trap of building things nobody needs. Virginia was determined not to fall into that trap. Every feature, every interface change, every decision had to have genuine customer value behind it.

Hiring for Grit, Not Resumes

Building a mission-driven team in a niche industry like wastewater tech requires a different hiring philosophy.

Nyad AI ditched resumes entirely.

Instead, they look for three things: First, has someone done something extraordinary? It doesn’t matter what. One team member spent 40 days in Antarctica. Another can run a 220-mile marathon. The bar is just that they’ve done something that shows they push themselves.

Second, they look for high agency. People who take ownership and drive things forward.

Third—and this is the one that stuck with us—they look for what Virginia calls “the dog in them.” They’ve literally got a pair of janky Bluetooth speakers in the office that are ridiculously difficult to connect. When someone visits, they ask them to connect them. Do they give up after 90 seconds? Or do they keep going like a dog with a bone until it’s done?

It’s a proxy for persistence, problem-solving, and that hunger to get things done.

What’s Next for Nyad AI

The vision for Nyad AI is ambitious. They’re building what Virginia calls “a Jarvis for wastewater”—a system that will eventually integrate with all the other data infrastructure at a treatment plant, providing comprehensive operational intelligence and recommendations for process control changes.

But they’re not there yet, and that’s intentional. They’re staying focused on the core value: helping operators quickly and accurately detect organisms in wastewater samples, turning that data into actionable insights.

It’s a reminder that transformative companies often start by solving one problem really well. The microscope-to-AI journey at Nyad AI isn’t just about technology—it’s about respecting the operators doing this critical work and giving them tools that actually fit into their world.

Want to hear Virginia’s full story, including how she navigated the transition from law school to water tech, what it’s like building a startup with your spouse, and the specific technical decisions that shaped Nyad AI’s product? Listen to the full episode of Code Story: Insights from Startup Tech Leaders.