Research
A research program built around one question: how can AI turn Earth observation data into timely, local decisions — starting with Nepal's farms, glaciers, and floodplains.
Research Interests
Deep learning architectures purpose-built for spatial and temporal Earth data.
Multispectral & SAR imagery analysis for land, water, and vegetation monitoring.
Spatial data infrastructure, geoprocessing, and web mapping.
Image classification and segmentation applied to satellite imagery.
Ensemble methods, cross-validation rigor, and model interpretability.
CNNs and transformer-based models for spatio-temporal prediction.
Sentinel, Landsat, and ERA5 data pipelines for continuous monitoring.
Climate risk indicators and long-term environmental trend analysis.
Current Research
Comparing SVR, Random Forest, Gradient Boosting, and XGBoost on multi-stage NDVI and canopy temperature data for wheat yield estimation, with emphasis on rigorous cross-validation and generalization across growth stages.
Extracting NDVI, ERA5 climate, and Copernicus land-cover datasets over Nepal's bounding box via Google Earth Engine for open, reusable research datasets.
Future Research
Research Goals
Publish first preprint; finalize scholarship applications.
Graduate B.E.; begin fully funded MSc/PhD program.
Publish peer-reviewed GeoAI research; contribute to open datasets.
Lead applied GeoAI research for climate resilience in South Asia.
Research Output