Database, AI, and Spatial Systems (DAISS) Lab

The Database, AI, and Spatial Systems (DAISS) Lab focuses on developing AI-based frameworks for geospatial and spatiotemporal datasets. The lab also works on optimizing database systems and AI systems, utilizing the best of both domains. Focus areas include Geospatial AI, Geospatial Data Analytics, Machine Learning Systems, and Database Systems.

Lab Members

• Kanchan Chowdhury
  • Role: Lab Director
• Samya Acharja
  • Role: Graduate Research Assistant and PhD Student
  • Fall 2025 - Present

Focused Projects

The projects of DAISS Lab can be summarized as following:

• Spatiotemporal Representation Learning: learning vector embeddings for geospatial and temporal datasets focusing on capturing geospatial object geometry, topology, and contextual semantics. Target applications include but are not limited to disaster management, emergency response, and transportation management.
• AI for Databases: utilizing machine learning techniques for automatic optimization of geospatial database query engines and translating natural language questions into geospatial database queries.
• Database for AI: co-optimizing end-to-end data science pipelines comprising data preprocessing and machine learning inference, mainly focusing on In-Database ML systems.
• AI for Raster Imagery: application oriented research combining satellite imagery and AI to solve real-world problems in the domain of agriculture and natural disasters.

Publications

Listed here

Past Projects

• GeoTorchAI Deep Learning Framework
  • GeoTorchAI, formerly known as GeoTorch, is a spatiotemporal deep learning framework on top of PyTorch and Apache Sedona. It enable spatiotemporal machine learning practitioners to easily and efficiently implement deep learning models targeting the applications of raster imagery datasets and spatiotemporal non-imagery datasets. Besides deep learning, it also supports scalable and distributed data preprocessing for raster and spatiotemporal datasets.
  • Project Link: Github Repository
• InferF: Model Decomposition Framework
  • An out-of-box AI/ML-SQL co-optimization approach for end-to-end inference workflows where the users specify a SQL query and a pre-trained model exported in ONNX format, and the end-to-end processing will be automatically optimized reducing the execution latency.
  • Project Link: Github Repository
• ML Aware Spatial Data Repartitioning
  • This is a framework which aims at reducing the training time and memory usage of a spatial machine learning model by reducing the number of partitions in a spatial grid dataset. Experiments on four datasets achieved significant reduction in training time and memory consumption while bounding the difference in prediction error within 5%.
  • Project Link: Github Repository