Geospatial data scientist specializing in hyperspectral & SAR analysis, crop-yield machine learning, and production EO pipelines FastAPI · GEE · STAC. M.Tech, IIT Guwahati · presented at EGU 2024.
$ whoami
geospatial-data-scientist · M.Tech IIT Guwahati
$ cat focus.txt
hyperspectral (PRISMA) · SAR (Sentinel-1) · crop-yield ML
$ ls pipelines/
spectral-indices time-series lulc flood-sar
sebal-et aef-embeddings landslide-risk
$ ./deploy --stack fastapi,stac,gee
✓ 15+ endpoints live · 7 analysis modules
$
I am a Geospatial Data Scientist (M.Tech, Earth System Science, IIT Guwahati) with a Civil Engineering foundation. My PRISMA hyperspectral thesis on hydrocarbon microseepage was presented at EGU 2024.
Currently at SwanSAT, I build the analytics backend of a crop-monitoring platform — multi-temporal satellite analysis (SAR, Optical, Thermal) powering yield forecasting, flood detection, and landslide risk assessment.
I also bring expertise in 1D/2D hydraulic modelling, flood risk assessment, and hydrology — combining traditional engineering principles with modern Python-based geospatial tooling.
Hyperspectral, multispectral & SAR acquisition across PRISMA, Sentinel and Landsat missions.
Ensemble ML, spectral unmixing, foundation-model embeddings, energy-balance physics.
Production FastAPI services, STAC pipelines and interactive dashboards used for real decisions.
From a production crop-intelligence platform to open geospatial tools — everything I've shipped.
Production crop-monitoring platform — I built the analytics backend: 15+ FastAPI endpoints across 7 analysis modules on Microsoft Planetary Computer STAC imagery, served to a React + Leaflet dashboard.
NDVI, GNDVI, NDWI, NDMI, LST plus custom band math with interactive stretch & statistics.
Cloud-filtered mean/min/max/std crop-trend curves over any date range.
ESA WorldCover 10 m with automatic per-class area breakdowns (km²).
Sentinel-1 VV/VH, Lee despeckling, dB-drop thresholding, before/after flood masks.
Landsat thermal/optical energy balance with automated ERA5-Land anchor-pixel calibration.
K-Means clustering & cosine-similarity search over a 64-band embedding index.
All-India district statistics plus corridor exposure for 1,000+ National Highways (250 m–1 km buffers, per-class length shares).
Some demos run on Streamlit Community Cloud and may hibernate. If you see a "Zzzz" screen, click "Yes, get this app back up!" — it wakes within moments.
Full-stack geospatial web app on Streamlit + GEE: real-time Vegetation Health Index monitoring, soil-moisture retrieval and flood mapping with ML/DL tooling.
Serverless geospatial ETL: automated conversion between Shapefile, GeoJSON & KML with CRS reprojection handling for clean interoperability.
Satellite-driven site selection for sustainable water conservation — DEM, SAR & optical fusion to locate optimal rainwater-harvesting structures.
Lightweight, responsive WebGIS module for visualizing custom datasets and layers with a seamless map-first user experience.
Interactive dashboard for landslide risk assessment along Indian National Highways using InSAR ground-deformation data, district-level susceptibility mapping, and corridor exposure analysis.
District-level climate anomaly analytics dashboard across India — tracking temperature, precipitation, and vegetation anomalies with interactive maps and time-series visualizations.
SwanSAT · Mumbai, India
Indian Institute of Technology Guwahati
Rajkiya Engineering College, Ambedkar Nagar
Kumar, N., Ahmad, A., Pal, A. K., Nair, A. M.
Satellite image-analysis approach for mapping hydrocarbon microseepage anomalies, validated against field-spectrometer and geochemical ground truth.
Read Paper