Spaces:
Sleeping
Apply for a GPU community grant: Personal project
I’m building NEPA.tools AI, an open-source Hugging Face Space that makes U.S. environmental review data easier to access for the public, researchers, and journalists.
The app uses the PNNL/NEPATEC2.0 dataset (about 61,881 projects, 142,083 documents, and ~7M pages) and lets users:
Ask natural-language questions about NEPA projects
Search/filter by agency, location, category, and sector
Get grounded, project-level answers from retrieved records
Technically, the Space combines Gradio + Transformers + Datasets with hybrid retrieval (semantic + keyword) and a RAG chat layer. The codebase is open and being actively improved for reliability, transparency, and reproducibility.
I’m requesting a free or discounted HF compute tier to stabilize inference and scale community use. Right now, ZeroGPU availability can intermittently fail during peak demand. Grant support would be used to provide reliable uptime and continue open development.
Planned open contributions:
Public repo updates and documentation
Retrieval quality benchmarks/evaluation notebook
Better citations/export features for users
API-ready integration with the NEPA.tools ecosystem
This project is a side project with public-interest impact: improving transparency and discoverability of environmental permitting data for everyone.