Research & projects
Research projects and case studies
Three projects that show how I work: field and laboratory data, GIS and remote sensing, and machine learning applied to real rivers in Germany and Bangladesh.
Jump to: AI-Driven River Feature Inventory Agent · Erft River hydromorphology · Bhairab River water quality
Mar 2026 – present · Independent research project · Cologne, Germany
AI-Driven River Feature Inventory Agent
Drone-based hydromorphological mapping with YOLO and ByteTrack
Context
Instream wood — fallen trees, leaning trees, log jams and stumps — matters for river restoration and for flood-risk monitoring. Counting it from drone footage is slow by hand, and the same object appears in many overlapping frames, so naive detection double-counts.
Approach
- YOLO object-detection backbone to detect instream wood in drone imagery and video.
- ByteTrack persistent-ID tracking, so each object keeps one identity across frames and is counted once.
- Every detection georeferenced to WGS84 coordinates using UAV telemetry and camera intrinsics.
- Agent-in-the-loop pipeline built in Google Antigravity, orchestrated with Claude Code and a custom YOLO MCP server.
Output
A deduplicated, georeferenced inventory of instream wood, exported as GIS-ready JSON and Excel files for hydromorphologists.
Tools & methods
Sep 2025 – Feb 2026 · MSc coursework · TH Köln
AI-Supported Hydromorphology Mapping & Habitat Assessment — Erft River
Drone imagery, machine learning and a LAWA-OS field survey
Context
A hydromorphological assessment of the Erft River for my M.Sc. at TH Köln, combining remote sensing and machine learning with a conventional field survey.
Approach
- Drone imagery and GIS for spatial coverage of the river corridor.
- Image annotation in Supervisely and Roboflow, and machine-learning models — YOLOv8, RT-DETR and a MATLAB CNN — to identify hydromorphological features.
- Field-based river habitat assessment with the LAWA-OS framework and RiverVision, documenting channel morphology, bank structure, substrate and vegetation.
Output
Results compiled into an interactive ArcGIS dashboard — explore it below.
Tools & methods
Jan 2024 – Mar 2025 · Research project · Khulna, Bangladesh
Water Quality & Hydrological Assessment of the Bhairab River
GIS, Python and remote sensing in Khulna, Bangladesh
Context
A study of the Bhairab River in Khulna, Bangladesh, assessing water quality and the river’s capacity for self-purification under anthropogenic pressure.
Approach
- Water quality assessed through pH, TDS, EC, BOD, COD and dissolved oxygen.
- River mass balance and self-purification capacity evaluated with correlation and regression methods.
- Water Quality Index (WQI) mapped spatially in QGIS and ArcGIS.
- Google Earth Engine land use / land cover (LULC) analysis to assess anthropogenic impacts.
- Statistical and multivariate trend analysis in Python (pandas, matplotlib, scipy).
Output
Findings published as peer-reviewed research in the Journal of Civil Engineering and Materials Application (March 2025). See publications.
Tools & methods
Earlier research
My B.Sc. thesis (2023) assessed surface water quality in the Dharatana River, Bagerhat, using the Water Quality Index — later published as Assessment of Surface Water Quality Using the Water Quality Index (WQI): A Case Study of the Dharatana River, Bagerhat. Further studies on river water-quality parameters, domestic wastewater and construction-waste management are listed under publications.
Get in touch
Working on water quality, wastewater or river monitoring?
I’m looking for internship and Werkstudent positions in Germany, and I’m happy to talk about research collaboration. Email or LinkedIn is the quickest way to reach me.