
AI-Powered Geospatial Assistant for Smart Cities
Synopsis
LAMP (a Language Model on the Map) technology integrates advanced language models with geospatial data to deliver intelligent, context-aware responses to location-based queries. This AI solution enhances urban planning, navigation and decision-making, offering tailored applications for building smart city.
Opportunity
At smart cities evolve, the demand for intelligent systems capable of handling complex, location-specific queries continues to grow. General-purpose language models, while proficient in natural language processing, lack the spatial awareness and localised knowledge essential for urban planning and geospatial applications.
LAMP (a Language Model on the Map) addresses this gap by combining the capabilities of large language models with comprehensive geospatial datasets. It provides accurate, context-aware insights for diverse urban stakeholders, including city governments, urban planners, tourism boards and businesses operating in urban areas. The technology’s versatility spans a range of use cases, from improving daily commutes for residents supporting strategic infrastructure development. As cities worldwide prioritise efficiency and smarter urban services, LAMP technology offers a transformative solution for enhancing city operations and citizen experiences.
Technology
LAMP leverages a groundbreaking approach that integrates large language models with detailed geospatial data. By fine-tuning on city-specific information—such as points of interest, addresses, opening hours and geographical positions—it delivers highly relevant responses tailored to urban needs.
Its self-supervised training process enables the model to understand spatial relationships and respond to complex natural language queries. LAMP's capability for spatial reasoning allows it to interpret proximity, plan multi-stop routes and provide nuanced recommendations. These features empower users with actionable insights, from personalised navigation to context-aware urban planning decisions.
-the-training-phase-to-inject-the-geospatial-knowledge-into-the-language-model-(right)74c60480-091d-43a3-8e26-1e78a20c28c7.png?sfvrsn=3afeb52d_1)
Figure 1: [Left] The synthetic conversational data generation process. [Right] The training phase to inject the geospatial knowledge into the language model.

Figure 2: Example response of our Language Model, when asked to plan a trip. The response is consistent with the user’s query and current location.
Applications & Advantages
- Tourism: Personalised city exploration and tailored recommendation systems
- Urban planning: Advanced decision support for infrastructure development
- Emergency services: Rapid identification of nearby facilities for emergencies
- Retail and hospitality: Location-based customer engagement and analytics
- Public services: Enhanced access to city resources and real-time information
- Real estate: Comprehensive neighbourhood analysis and property valuation