PhD Candidate · Purdue University · Hydraulics & Hydrology
I build AI-Native Hydrology — an open ecosystem of agent-operable Python packages and an MCP server that lets a general AI agent conduct end-to-end hydrological research with full scientific defensibility as a structural byproduct.
My PhD at Purdue frames this across three pillars: Diagnose (what breaks when AI meets hydrology), Fill the Gaps (build the missing infrastructure), Democratize (open tools anyone can extend).
github.com/AI-Hydro · agent-native · Apache 2.0
A layered family of purpose-built packages, all surfaced through aihydro-tools as the MCP server meta-package:
| Package | Role | Links |
|---|---|---|
aihydro-tools |
MCP server · 144 validated tools · AI agent surface | PyPI · Zenodo |
aihydro-modelling |
Coming soon | — |
aihydro-watershed |
Basin delineation + hydrological signatures · anywhere on Earth | PyPI · Zenodo |
aihydro-lsh |
Coming soon | — |
aihydro-data |
Global data router · 54 products · 8 regions | PyPI · Zenodo |
aihydro-core |
Zero-dep substrate · HydroResult contract · bootstrap CI | PyPI · Zenodo |
camels-attrs |
CAMELS-US attributes for any USGS gauge | PyPI · Zenodo |
pygeoglim |
Global geology attributes · GLiM/GLHYMPS · area-weighted | PyPI · Zenodo |
- swatplus-builder — agent-native SWAT+ model construction with 7-gate claim governance
- PhD research in progress — two science pillars in hydrological modelling and global datasets (publishing 2026–2027)
large-sample hydrology · defensible AI · foundation models for Earth systems · agent-native scientific computing · MCP protocol · differentiable modelling
Purdue University · Lyles School of Civil & Construction Engineering · Advisor: Dr. Venkatesh Merwade