“If you sit in one place long enough, the world starts to tell you where you are.”
That quiet insight has followed Dr. Nathan Jacobs ever since graduate school. During his Ph.D. research at Washington University in St. Louis, he studied “passive vision” — how much a computer could learn simply by watching. His lab collected thousands of webcam images from coffee shops, ski slopes, and beaches around the world. Over time, the pictures revealed their own geography: clouds drifting, light shifting, seasons turning.
“What began as a vision problem,” he recalls, “slowly became a fascination with place itself – how change tells a story.” That curiosity still anchors his work today, as he helps WashU and the Taylor Geospatial Institute connect data, people, and ideas across disciplines.
A Researcher ‘Watching the World’
Jacobs’ love of information goes back to the early days of the internet. “It was the wild west,” he says with a grin. “People were building brand-new search engines every week. I just wanted to understand how we find meaning in this growing sea of data.”
That drive led him to WashU to explore how computers see. His advisor’s project on “passive vision” – learning from ordinary webcam feeds – opened a new world. “I wasn’t a remote-sensing person who moved into AI,” Jacobs says. “I was a computer-vision person who became fascinated by location.”
As the web matured, he began experimenting with new sources of imagery, including social media photos, maps, and crowdsourced data. “There are all sorts of geotagged things that tell you about the world,” he says. “You just have to look at them long enough.”
Building a Geospatial Home
When Jacobs returned to WashU in 2022, he found geospatial expertise scattered across departments. “Everyone was doing something with data and maps, but we needed a way to see ourselves as a community,” he recalls.
Out of that effort emerged the Geospatial Research Initiative, a two-year program funded by the provost’s office to connect researchers across disciplines and clarify WashU’s unique strengths in geospatial science. Jacobs co-leads the initiative with Dr. Alex Bradley from Earth, Environmental, and Planetary Sciences, working closely with Dr. Michael Frachetti and Dr. Jennifer Moore, who serve as co-leads on the GRI leadership team alongside colleagues in anthropology and data services.
“We’ve run two rounds of internal seed grants to identify who’s doing what. It’s about understanding strengths, building infrastructure, and keeping it sustainable,” he says.
Building AI Tools to Transform Data into Research Insights
In recent years, Jacobs’ curiosity has taken him into new terrain: biodiversity and ecological modelling. His recent paper, TaxaBind: A Unified Embedding Space for Ecological Applications, which has sparked significant attention for its multimodal approach to ecological modelling, grew out of a deeply personal source of inspiration – his father’s lifelong work with the Missouri Department of Conservation.
TaxaBind builds on Jacobs’ broader research into multimodal learning for ecological and geospatial applications, work recently featured by the McKelvey School of Engineering in “Multimodal AI Tool Supports Ecological Applications”. The framework combines six information streams – from satellite and environmental data to text, sound, and species imagery – to support large-scale ecological modelling.
When Jacobs discovered iNaturalist, the citizen-science platform of geotagged wildlife photos, “it fit perfectly with everything I’d been doing with social media,” he says.
One ongoing project with the Living Earth Collaborative, led in partnership with Dr. Jonathan Losos and postdoctoral researcher Dr. Elizabeth Carlen, examines how sampling bias shapes ecological datasets, such as why iNaturalist records show fewer squirrel observations in some St. Louis neighborhoods.
“There are far fewer iNaturalist observations of squirrels in North St. Louis,” Jacobs notes. “It’s not that the squirrels aren’t there; there just aren’t as many people taking pictures of them.”
“The goal isn’t to replace ecologists,” he adds. “It’s to give them better tools. The real progress happens when we close the loop between domain experts and AI researchers.”
Beyond ecology, Jacobs’ team keeps stretching what geospatial AI can do. Their RANGE framework blends imagery and neural-field encoders to capture subtle spatial context. More recently, projects like Sat2Cap and GeoSynth, featured in “Artificial Intelligence Meets Cartography”, use text prompts to generate realistic satellite scenes.
“I like sitting with people who see the world differently. An ecologist, a computer scientist — we speak different languages, but once you start listening, you realize AI is mostly about communication,” he says.
Democratizing Geospatial AI
As geospatial AI tools grow more powerful, Jacobs is also thinking about how to make them accessible. He sees promise in the rise of large language models and chat-based interfaces, which can let non-experts interact directly with data.
“Traditionally, remote sensing and spatial analysis have centered around policymakers or expert users who spend a lot of time with data. But we’re at a point now where we can start democratizing that, making it easier for a much wider range of people to ask their own questions and get meaningful results,” he says.
That same spirit of accessibility extends into his applied collaborations. At both WashU and TGI, Jacobs focuses on translating research into action, from biodiversity modelling to search-and-rescue. His collaborations with Yevgeniy Vorobeychik produced Visual Active Search and “Meta-learning to Find Every Needle in a Haystack”, systems that use reinforcement learning and aerial imagery to guide real-world searches. “The world looks different in different places,” he says. “Our tools need to adapt just as quickly.”
Earlier this year, Jacobs was appointed Assistant Vice Provost for Digital Transformation at WashU’s Digital Intelligence & Innovation Accelerator, a role that bridges research and infrastructure in the service of democratizing geospatial data.
He shrugs off titles, though. “What keeps me going is the sense of discovery,” he says. “The geospatial world is evolving fast, but at its heart, it’s still about understanding where things are, how they change, and why it matters.”
