PhD Candidate · Hydraulics & Hydrology · Purdue University
I build AI-native infrastructure for hydrological science — tools that let researchers move from question to reproducible result through natural conversation with their computational environment. My work sits at the intersection of differentiable modeling, large-sample hydrology, and the emerging Model Context Protocol ecosystem, advised by Dr. Venkatesh Merwade in the Lyles School of Civil and Construction Engineering.
I'm interested in what happens when foundation models stop being chatbots and start being collaborators: orchestrating real scientific computation, recording provenance automatically, and adapting to how individual researchers actually think and work.
An open platform where AI agents do real hydrological research — watershed delineation, streamflow analysis, model calibration, and reproducible provenance — all from a natural language conversation. Built as a VS Code extension on the Model Context Protocol, with a Python toolkit distributed via PyPI and a plugin architecture designed for community extension.
→ Documentation · VS Code Marketplace · PyPI
camels-attrs and pygeoglim — published Python packages for catchment attribute extraction and geological characterization, both with archived Zenodo DOIs and used in ongoing large-sample hydrology work.
- Differentiable and physics-informed hydrological modeling
- Large-sample hydrology and next-generation CAMELS-style benchmarks
- Reproducibility and automated provenance in computational Earth science
- AI-native scientific workflows and the Model Context Protocol
- National-scale SWAT calibration and watershed characterization
- Community-extensible research infrastructure
Python · PyTorch · TypeScript · FastMCP · MCP · SWAT · USGS APIs · HyRiver · xarray · GeoPandas · Taichi
AI-Hydro Org · Documentation · YouTube · PyPI
Purdue University · Lyles School of Civil and Construction Engineering · Apache 2.0