Artificial Intelligence for Land and Survey is an AI assistant launched by Eyad Al-Naimi and Naser Diabat.
The project aims to assist Jordanian citizens by answering their questions about the Department of Land and Survey. It also showcases the team’s skills in building AI solutions using limited tools and resources.
🔑 Key Techniques
-
Semantic Caching – speeds up responses by retrieving frequent questions directly from the FAQ knowledge base.
-
Router LLM – decides whether a user query belongs to the FAQ or requires searching the extended knowledge base.
-
Reranker – ensures only the most relevant documents are selected.
-
Matryoshka Embeddings – optimize search efficiency while minimizing information loss.
-
Prompt Engineering – ensures the model behaves as instructed.
-
LLM via Perplexity API – a large, fast model is used to provide users with the most accurate and high-quality answers possible.
🎯 Goals
-
Provide reliable and quick answers to citizen inquiries.
-
Demonstrate AI innovation with limited resources.
-
Highlight practical applications of retrieval-augmented generation (RAG) techniques.
📊 Data Source
- The knowledge base was built using data extracted from the Legislations section of the official website of the Department of Land and Survey.