GDLM - Geospatial Data Language Model

Short description of the innovation study

Geospatial data is critical for industrial decision-making, but accessing and integrating it remains complex and requires specialised expertise. Building on Geo-Llama from FFplus Call 1, DataMonkey will develop GDLM, a reasoning-based Geospatial Data Language Model designed to close the reliability gap between technical proof of concept and industrial deployment. Using European HPC infrastructure, advanced reasoning models, full-parameter fine-tuning, reinforcement learning, and a proprietary dataset extended with industry-specific training data, the team will improve geospatial query accuracy for real-world workflows. The goal is to deliver reliable general and industry-specific GDLM variants for insurance, mobility, and energy, enabling companies to integrate geospatial intelligence faster and at scale.

Organisations involved: 

Main SME: Urban Monkeys GmbH