I build production-grade data pipelines, ML systems, and geospatial analytics products where satellite data meets real-world decisions. My work sits at the intersection of SAR/InSAR remote sensing, scientific software engineering, and applied AI.
Currently at Brockmann Geomatics Sweden AB - working on Horizon Europe OASIS and ESA AQUATIME projects.
| 🛰️ EO data pipelines | End-to-end ingestion, preprocessing, validation, and delivery for multi-TB satellite imagery datasets (SAR, optical, multispectral) |
| 🤖 Geospatial ML systems | Classification, change detection, anomaly detection, and time-series forecasting on Earth observation data |
| 🔍 AI retrieval infrastructure | RAG pipelines, vector search (FAISS, ChromaDB), and multimodal retrieval for geospatial and document workflows |
| ⚙️ Reproducible scientific workflows | Modular, validated, and traceable pipelines for operational and research applications |
| Project | What it does | Stack |
|---|---|---|
| Disaster GeoRAG 🏆 2nd place · BiDS 2025 | Retrieval-augmented VLM pipeline for EO imagery triage in disaster response | Python · FAISS · LangChain · VLM |
| Vision Text Extractor | Privacy-aware OCR CLI - local (SmolVLM, LLaVA) or cloud (GPT-4o), one-command setup | Python · pixi · HuggingFace · Ollama |
| Similarity Search - ChromaDB | Semantic search system with FastAPI backend, web dashboard, CI pipeline, and test coverage | Python · ChromaDB · FastAPI · SentenceTransformers |
| LiDAR Data Processor | Point cloud processing and sampling for forest structure analysis | Python · LiDAR |
| InSAR / pygmtsar | DInSAR and PSInSAR time-series workflows for ground deformation monitoring | Python · GMTSAR · Sentinel-1 |
domains = ["SAR/InSAR/PolSAR", "EO time-series", "Geospatial ML", "GeoAI", "LiDAR"]
languages = ["Python", "SQL", "Bash"]
geo_stack = ["GDAL", "Rasterio", "GeoPandas", "xarray", "QGIS", "Google Earth Engine"]
ml_stack = ["scikit-learn", "PyTorch", "MLflow", "FAISS", "ChromaDB", "LangChain"]
infra = ["Docker", "Git", "GitHub Actions", "FastAPI", "pixi", "Linux", "LUMI HPC"]| Project | Goal |
|---|---|
| 🔄 Geospatial vector retrieval for SAR change detection | Production-ready GeoAI retrieval combining spatial indexing with semantic search |
| 🔄 EO time-series forecasting - transformer + classical hybrid | Benchmarking PatchTST, N-HiTS, and Mamba on environmental sensor data |
| 🔄 AQUATIME / OASIS pipelines @ Brockmann | Reproducible ML workflows for water quality and land surface analytics at ESA/EU scale |
upskilling = {
"data_engineering" : ["Apache Airflow", "dbt", "Kafka", "GeoParquet", "PostGIS"],
"cloud_infra" : ["Kubernetes", "Terraform", "GCP (BigQuery, GCS)", "AWS S3/Lambda"],
"mlops" : ["model registries", "feature stores", "pipeline orchestration"],
"geospatial_new" : ["COG workflows", "STAC catalogues", "OpenEO", "Zarr at scale"],
}7 peer-reviewed publications in Earth observation and geoscience - SAR tomography · PolSAR calibration · InSAR time-series · land subsidence · UHI analysis
| Organisation | Role | Period |
|---|---|---|
| Brockmann Geomatics AB | Geospatial Data Scientist | 2026 – present |
| ICEYE Oy | SAR Remote Sensing Engineer | 2021 – 2026 |
| TU Delft | Remote Sensing & GIS Researcher | 2020 – 2021 |
| IIT Bombay | Jr. Remote Sensing Researcher | 2019 – 2020 |
| ISRO / IIRS | M.Eng. Remote Sensing & GIS | 2017 – 2019 |
LinkedIn · Portfolio · Google Scholar · udit.asopa@gmail.com
| Category | Language & Tool |
|---|---|
| Languages & scripting |      |
| Geospatial & Earth observation |    |
| ML & data science |         |
| APIs, backends & data services |     |
| Cloud & infrastructure |       |
| Dev environment & tools |     |