Meet: Dr Eric Orenstein

Eric Orenstein is an AI+ Fellow at King's College London. These interdisciplinary fellowships are central to King’s £18 million strategic investment in academic excellence and demonstrate King's commitment to transforming AI research and innovation across all disciplines.

Initially, Eric didn't set out to become an ocean scientist. He grew up in Washington DC and studied physics at a liberal arts college of around 2,000 students. Everything changed when his mentor pushed him to do an internship at the Scripps Institution of Oceanography at UC San Diego.

He got a place on a research vessel in the Gulf of California for three and a half weeks. "Suddenly a world of adventure, cool toys, and scuba diving for work opened up," he says. He went back to Scripps to do his doctorate and never looked back. His PhD work centred on building an in situ microscope, a device moored near the beach that watched a fixed patch of ocean continuously. Over five years, it generated around a billion regions of interest, tiny images of plankton and microorganisms drifting past the lens. Nobody could look at them all. That bottleneck pulled Eric toward machine learning, splitting his time between Scripps and the electrical and computer engineering department at UCSD. From Scripps, he moved to the Monterey Bay Aquarium Research Institute to work on marine robotics, then to the National Oceanography Centre in Southampton, and now brought his research to King's College London.

The ocean, Eric explains, is a challenging space for the kind of AI most people build. Biological records exist for only around five percent of the ocean’s volume concentrated near shipping lanes, islands, coastal zones, and the surface waters of the northern hemisphere. Climate change makes things harder still, as populations shift and ecosystems reorganise. Now, his current focus is adaptive sampling: building AI for autonomous underwater vehicles that can decide what to do next in real time, with no human in the loop and no way to call home for extra computing power. Think of a self-driving car, operating in a mostly unknown environment, under 200 metres of water.

One capability he's developing is what computer scientists call novel category discovery. A system learns to recognize that something in front of it is different from anything it has seen before. The vehicle slows down and gets a closer look, as it might be an animal no catalogue has ever recorded. "When you're there," Eric says, "sometimes something new or different happens that you won't be able to repeat."

Eric is a co-creator of FathomNet, an open-source database of annotated marine images built at MBARI with collaborators from CVisionAI and the Ocean Discovery League. Ocean science has historically been protective of its datasets, partly because collecting them is expensive and slow. FathomNet tries to shift that culture. "No single institution, however well-resourced, has all the data needed to train models that do exactly what they want to do," he says. Since launch, submissions have come in from universities in Hawaii and institutions along the coast of British Columbia.

In his diverse career, Eric has operated small vessels, run field programmes, and collected data in conditions that most AI researchers have never encountered. That physical experience shapes the way he frames problems. "The problems we meet in the field are different from the ones most AI systems are designed for," he says. You can't see that difference from a dataset, you have to get wet.

He came to King's because he missed the cross-disciplinary contact and King’s offers a unique, diverse research space. Getting data to decision-makers at organisations like DEFRA requires working with people who think about human-computer interaction, science communication, and public policy. "You need to be the human face of this," he says. "It can't just be a number on a page."

Eric hopes to take students into the field, because he believes you can't solve a problem you've never stood in front of. From his office in central London, he still holds an affiliation with the National Oceanography Centre, which manages much of the UK's national marine observating infrastructure. He's watching for a window to get back out on the water. The ocean keeps moving and he plans to keep up.