Applied GeoAI & AI Systems
I work where emerging AI capabilities meet real geospatial problems.
My focus is not only on developing models, but on what it takes to turn them into reliable geospatial workflows: preparing data, evaluating performance, integrating with platforms, designing reproducible processes, and understanding how results support real decisions.
I am particularly interested in the connection between:
Data → Models → Evaluation → Workflows → Decisions
This systems perspective shapes how I approach applied AI. Model performance matters, but so do data quality, uncertainty, reproducibility, usability, and the context in which the technology is ultimately applied.
My current work at Esri Canada spans three areas.
- GeoAI for Imagery: I develop and evaluate computer vision workflows for satellite, aerial, drone, and enterprise imagery. Applications include object detection, image classification, semantic segmentation, change detection, and geospatial foundation models. Beyond model performance, I focus on how these workflows handle real-world variation in imagery, data quality, scale, and operational requirements. A major focus is making these capabilities usable within established GIS workflows rather than leaving them as standalone research prototypes.
- Geospatial Foundation Models: I work on integrating emerging foundation models with operational geospatial environments, including model configuration, metadata, integration testing, packaging, and workflows for Earth observation analysis. A key part of this work is understanding how these models perform across real-world datasets, resolutions, geographies, and analytical tasks; not just controlled benchmarks. This work explores how increasingly general AI models can become practical tools for spatial analysis.
- Generative & Agentic AI: I also explore how LLMs, retrieval-augmented generation, AI assistants, and agentic systems can change the way people interact with geospatial technology. The interesting question is not simply whether an AI agent can call a tool. It is how AI systems can help people discover, execute, explain, and repeat complex spatial workflows reliably. I am particularly interested in tool orchestration, retrieval and grounding, workflow planning, human oversight, and the reliability of multi-step geospatial reasoning.
Selected Projects & Research Partnerships
Marine Mammal Monitoring with AI
Fisheries and Oceans Canada
Supported applied GeoAI work for detecting beluga whales in aerial imagery and communicating the resulting deep learning workflow. The work demonstrated how computer vision can support marine mammal monitoring across large geographic areas and was featured in one of Esri's ten most-read stories of 2025.
Ongoing collaboration is supporting graduate research extending customized deep learning approaches to additional marine mammal species.
Read the Esri story - Listen to the podcast - View the StoryMap
Ocean Acidification Mapping in Canada
Fisheries and Oceans Canada, MEOPAR, Canadian universities & research organizations
Contributed to a national initiative connecting ocean acidification science with geospatial analysis, data accessibility, mapping, and knowledge mobilization.
The collaboration resulted in a national peer-reviewed assessment published in Frontiers in Marine Science and supports broader efforts to make ocean acidification evidence more accessible for research and decision-making.
Adaptive Wildfire Evacuation
University of Calgary
Support a research project investigating AI-powered adaptive route optimization for wildfire evacuation. The work combines GPS movement data, reinforcement learning, GIS, and interactive visualization to explore how evacuation routes can adapt as wildfire conditions and transportation constraints change.
Geospatial Machine Learning Inside SQL
University of New Brunswick
Support research investigating how geospatial machine-learning operations can be represented as relational algebra operators and integrated directly into standard SQL workflows. The research explores a larger question for applied AI: how advanced spatial analytics can become accessible inside the data systems organizations already use.
Accurate Forest Carbon Quantification
York University
Contribute to ongoing research with York University focused on improving the accuracy of forest carbon quantification through geospatial analysis, remote sensing, and the application of Esri technology. The project supports broader efforts to strengthen forest carbon monitoring, assessment of carbon dynamics, and nature-based climate solutions in Canada, while exploring more scalable and reproducible approaches for integrating spatial data into carbon accounting workflows.