Skip to content

Spatial Data Science & Machine Learning

I am a Geocomputational Scientist and Geosensing Engineer specializing in Spatial Data Science, Remote Sensing, and Machine Learning.

My work bridges physical geoprocesses, high-performance computing (HPC), and deep learning. I develop automated pipeline and modeling solutions to process high-resolution Earth observations (optical, radar/SAR, LiDAR, and hyperspectral), and fuse them with land surface models to solve complex hydrological, glacial, and permafrost challenges.

Core Specializations

  • Geospatial & Remote Sensing

    • Active and passive sensor analysis (LiDAR, optical, SAR, hyperspectral).
    • Satellite data fusion (Landsat, Sentinel, SWOT, VIIRS, MODIS, ICESat-2).
    • Large-scale terrain modeling and land surface simulation.

  • AI/ML

    • Machine Learning and Computer Vision for river and water classification.
    • Transformer models and custom geospatial embeddings (e.g., Clay, Prithvi).

  • High-Performance Computing

    • Scale-out distributed workflow systems utilizing SLURM, MPI, and Dask.
    • Multi-GPU/Distributed Data Parallel (DDP) deep learning training.
    • Cloud integration (AWS, serverless Lambda execution).

Technical Stack

Category Technologies, Libraries, and Frameworks
Geospatial & GIS GeoPandas, Rasterio, Rioxarray, GDAL/OGR, Xarray, Laspy, Arcpy, QGIS, PostGIS, H3
Machine Learning & AI PyTorch, Scikit-Learn, XGBoost, Hydrology Transformers, PEFT, Ollama
Computation & Scale NumPy, Pandas, Dask, Joblib, Multiprocessing, Bash, HPC/SLURM, AWS Lambda
Visualization & UI Plotly/Dash, Matplotlib, Cartopy, hvPlot, Kepler.gl, Virtual Ice Explorer
Programming Languages Python, Julia, SQL, Matlab, R, C++

Key NASA & NSF Work

🌊 NASA Regional Hydrology & SWOT Mission

Developed machine learning algorithms and spatial pipelines to ingest and process SWOT (Surface Water and Ocean Topography) observations. Created interactive visualization dashboards (swotvis.cuahsi.io) to monitor surface water elevation and river width dynamics worldwide in support of international hydrology sciences.

❄️ Daily Snow Water Equivalent (SWE) of North America

Architected a large-scale data fusion pipeline on high-performance computers. We integrated NASA MODIS and VIIRS satellite snow observations with land surface simulation outputs, generating a pristine 20-year daily continential-scale Snow Water Equivalent grid at 0.01° high resolution.

🏔️ Glacier Velocity and Sentinel-2 Corrections

Co-developed empirical correction workflows to eliminate systematic orthorectification offsets from Sentinel-2 velocity fields. Applied these pipelines to map glacial ice stream dynamics across the outlets and margins of Greenland.

🧪 Automated Laser Elevation Processing

Coded open-source Python toolsets for ICESat-2 altimetry analysis to calibrate and correct systematic vertical biases in continental-scale ArcticDEM, REMA DEM, and WorldDEM terrain models.