50 Shades of Green Tape

Inspiration

The inspiration for "50 Shades of Green Tape" comes from the need for transparency and efficiency in clean energy deployment and environmental permitting. The team envisions a system where various stakeholders - from American citizens to developers, agencies, decision makers, and clean energy stakeholders - can easily access and utilize permitting data and environmental impact information at a local level.

What it does

This project is a RAG-based (Retrieval-Augmented Generation) system that provides geospatial insights for NEPATEC 1.0. It uses document embedding and similarity search to offer semantic and geospatial insights from a vast database of over 300,000 embeddings from 1,000 projects. The system allows users to query information about environmental impacts, permitting processes, and clean energy deployment at a county level.

How we built it

1. We used NEPATEC 1.0 as the base dataset
2. Document chunking for processing large texts
3. OpenAI Embedding for creating vector representations of documents and queries
4. Geopy.geocode for handling location data
5. Similarity search algorithms to find relevant documents and locations
6. Kepler for visualization of geospatial data

Challenges we ran into

While not explicitly mentioned, challenges likely included:

1. Processing and embedding a large volume of documents (over 300,000 embeddings)
2. Integrating geospatial data with textual information
3. Designing an efficient similarity search algorithm for quick retrieval of relevant information
4. Creating a user-friendly interface for various stakeholders with different needs

Accomplishments that we're proud of

1. Successfully processing and embedding over 300,000 documents from 1,000 projects
2. Creating a system that can provide both semantic and geospatial insights
3. Developing a solution that addresses the needs of multiple stakeholders in the clean energy and environmental permitting space

What we learned

1. Large-scale document processing and embedding techniques
2. Integration of geospatial data with text-based information
3. The complexities of the NEPA process and environmental permitting
4. The importance of data accessibility in decision-making for clean energy deployment

What's next for 50 Shades of Green Tape

1. Mapping the most relevant environmental impacts for every county based on similarity embedding research
2. Identifying federal and local agencies involved in the permitting process for each county
3. Conducting sentiment analysis of applications, including approved and unapproved ones, for each county
4. Potentially expanding the system to cover more regions or types of environmental data
5. Refining the user interface to make it more interactive and accessible for different stakeholder groups

Built With

• docker
• embeddings
• fastapi
• geopy
• kepler
• openai
• pgvector
• postgresql
• python
• react