There was a certain excitement in the way he talked about his work. He had built things, broken things, rebuilt them and somehow kept inventing, like a true inventor does. There was the high-voltage plasma system made partly from electronic waste, then an AI-based system that could rapidly test water quality.

The question he had back then was fairly simple: Can I build something that works?

A year later, the question has changed.

Now, Zabeer is asking what happens when the thing he built leaves his table, enters the real world and has to survive someone else’s scrutiny.

That change is at the heart of UrbanSentinel, the system he has developed around water contamination. It is no longer just about cleaning polluted water or detecting that something is wrong. The idea is to build a system that can detect contamination, predict what might happen, trace where it may have come from, treat it and then check whether the treatment actually worked.

It sounds almost absurdly ambitious when put that way.

But the interesting part of speaking to Zabeer now is that he seems much less interested in making it sound impressive.

“My biggest change was that I stopped thinking about water pollution as only a treatment problem,” he said.

His earlier work had started with purification. But while working on it, he ran into another problem: even if you could clean contaminated water relatively quickly, figuring out exactly what was in it and whether the treatment had worked could take expensive laboratory equipment and, in some cases, days of testing.

That led to his AI-based rapid water-testing system.

Then came another problem.

A sensor could tell you that water was contaminated. But what good is that information if you do not know where the contamination came from, what direction it is moving in, or what you should do next?

That is where UrbanSentinel came from.

Zabeer describes it, rather nicely, as something like a nervous system for an urban water network.

The system uses several kinds of signals to examine water, including UV-visible absorption, infrared spectroscopy, Raman-based information and electrical properties such as capacitance and resistance. A lightweight deep-learning model then interprets those signals and estimates parameters such as BOD and COD.

At the network level, the ambition gets bigger. Data from different monitoring nodes could eventually be combined with information such as rainfall, water level and flow to predict whether contamination is likely to reach another point downstream.

Then comes the tracing.

If one node detects clean water and another detects contamination, the location and timing of those readings can be combined with the contaminant’s spectral fingerprint and a hydraulic model of the network to work backwards towards possible sources.

There is a very important caveat here, and Zabeer is quite deliberate about it. UrbanSentinel is not supposed to point at a factory and announce that it caused the pollution. It is meant to produce probability-ranked possible sources based on several pieces of evidence. Any actual enforcement decision would still require the appropriate regulatory and legal process.

That caution would probably have sounded less natural from the Zabeer I interviewed a year ago. Back then, much of his approach was about making something work. Now, working is only the beginning.

His time working with Buet’s Institute of Appropriate Technology has had a lot to do with that. He still builds things the way he always has, but he now thinks much more about calibration, control samples, reference methods, repeatability and uncertainty. In other words, he has learned that making something work once is not the same as proving that it works.

“Earlier, my approach was much more like that of an inventor,” he said. “I would identify a problem, build something, make it work and then improve it through repeated trial and error.”

Now there is another question at every stage: Where can this fail?

That became especially clear at the World Water Challenge 2026 in South Korea, where Zabeer had to defend UrbanSentinel in front of researchers and industry professionals.

The difficult part was not necessarily explaining what the system was supposed to do. It was explaining what it could actually do already.

The sensing system, the treatment system, the network-level source attribution and the future forecasting system are not all at the same stage. Some have experimental evidence behind them. Others are still being developed.

The questions about dataset size, scalability and how much computing could realistically happen on an ESP32-class device forced him to draw those lines more carefully.

One of the most important numbers in that conversation is also one of the least glamorous: 28.

That is the number of laboratory-paired samples in his current labelled dataset.

Across those samples, the sensing subsystem showed about 96.4% average agreement with reference laboratory results, while 92.9% of readings were within 10% of the corresponding laboratory value.

Those are encouraging numbers. But they are also not enough to declare that UrbanSentinel will work everywhere.

Zabeer knows that. The same honesty appears in the treatment results.

In one experimental dataset, the plasma treatment reduced colour from 468 to 115 Pt-Co and COD from 86 to 22 mg/L. Total coliform, faecal coliform and E. coli were reduced to zero in the tested samples.

But BOD went from 68 to 7 mg/L, which is certainly better, but still above the 5 mg/L target used in that analysis. The pH also fell from 7.2 to 6.4.

It would have been easy to focus on the impressive numbers and leave the rest somewhere at the bottom of the page. Instead, Zabeer now treats the limitations as part of the story.

He describes the plasma system more carefully as a polishing and disinfection technology within a treatment train, rather than claiming that it can replace every part of conventional wastewater treatment. That change in attitude may be more important than another technical feature added to the system.

“The strongest improvement is not adding another feature,” he said. “It’s narrowing a claim until the evidence genuinely supports it.”

There is something almost refreshing about hearing a young researcher talk this way. Especially because UrbanSentinel has already taken him quite far.

The project received an Outstanding Award at the 2026 World Water Challenge, putting a young Bangladeshi researcher in a room where the conversation was no longer simply about school projects and competitions. It was about evidence, feasibility, scalability and what happens when an idea has to work outside the environment in which it was created.

That is precisely what the World Water Challenge is built around: pushing water solutions beyond ideas and towards real-world implementation. Part of Korea International Water Week, the 2026 event was held in Daegu from 9–11 September under the theme “Water, Energy, AI: Securing Our Future”, drawing around 12,000 visitors.

Finalists were assessed on feasibility, sustainability, originality and impact before presenting their projects to an international expert panel. For Zabeer, the award definitely mattered, but perhaps more important was the fact that he was now being asked to defend not just what he had built, but what he could actually prove.

But the award is not where Zabeer thinks the story ends. If anything, it has made the next part harder.

He now wants to test UrbanSentinel in a controlled urban setting. His proposed first phase includes at least 500 laboratory-paired samples across 20 monitoring sites over 12 months, along with contaminant fingerprinting, nitrate and nitrite measurements, source-attribution testing and actual measurements of treatment energy consumption.

One possible testing area is the Turag/Gazipur–Konabari industrial corridor.

A successful field test, however, would mean much more than putting a few devices beside a river and waiting for a number to appear on a screen.

The nodes would have to keep working through rain, temperature changes, turbidity, electrical noise and sensor fouling. The AI predictions would need to be compared against hundreds of real laboratory samples. The network would have to detect contamination, follow its movement and narrow down its possible source.

And the treatment system would have to show that it can consistently reduce contaminants while accounting for energy use and possible by-products.

That is a much less glamorous stage of innovation. It involves maintenance, calibration, repeated testing, failed experiments, more samples and more waiting. It also involves people and institutions.

A real deployment would require researchers, water utilities, the Department of Environment, municipal authorities, industries and potentially development organisations to work together. Someone would have to maintain the nodes, someone would have to verify the results and someone would have to decide what happens when the system raises an alarm.

For a young researcher in Bangladesh, even getting to this point has its own complications.

Zabeer now has access to Buet laboratories and mentors, which has changed things significantly from the days when he was trying to build and test much of his work himself.

But specialised sensors, calibrated spectroscopy equipment and independent laboratory testing remain difficult to access. That problem, he says, has not disappeared. What has changed is his understanding of it.

He no longer sees laboratory access as something that would simply make his work easier. He sees it as part of the research itself. And perhaps that is where the biggest change in Zabeer lies. He still loves electronics. He is still fascinated by plasma, physics, AI and instrumentation. And he still wants to build things.

But when I ask him how he sees himself now, the answer is different from the one I might have expected a year ago.

He wants to become a researcher and engineer who can connect science with systems that solve actual problems, not someone whose work ends when a competition ends.

Then there is the question I liked most.

If he could go back and talk to the version of himself who was interviewed in 2025, what would he say?

“Do not be in such a hurry to prove that everything works.”

It is a strange thing for an inventor to learn. We usually celebrate the person who can make something work. We put the prototype on a table, take a photograph, give it a name and start talking about the future.

Research is less forgiving. It asks what happened when something did not work. It asks whether someone else can get the same result, whether it has been tested enough, and whether 28 samples are really enough to make a claim.

More importantly, it asks you to be honest about what you do not know yet.

Zabeer seems to have reached that uncomfortable place with UrbanSentinel. And perhaps that is the real story of the project.

It is not simply that a young researcher built a system that could one day help cities detect polluted water. It is that somewhere between the laboratory, Buet and an international panel in South Korea, he became less interested in proving that his research was successful and more interested in understanding what the evidence could actually support.

That is a much harder question. It is also a much more important one.