Open geospatial intelligence for public impact
Areas of impact
Transit, movement, and accessibility models for fast-growing cities, supported by satellites, crowdsourced data and participatory evidence.
Water monitoring, flood assessment, and decision support connecting hydrology with infrastructure.
Neighborhood-scale spatial models for heat, flooding, and defensible climate-adaptation decisions.
Crop representation learning, time series, and world models for agricultural planning and resilience.
Tech Focus
Learn transferable spatial features from unlabeled, multimodal remote-sensing data.
Embed physical structure into neural surrogates for reliable environmental simulations.
Reduce annotation effort by selecting the most informative locations and samples.
Connect spatial data, analytical tools, and human goals through agentic decision support.
Explore the work
Open tools and applied research developed with institutions, communities, and public systems.
A travel-demand estimation and network-design library being developed alongside bus-route design for Gurugram Metropolitan City Bus Limited (GMCBL). An extension supports design decisions for metro and ring-road projects in the Chandigarh Tricity area.
A water-quality and quantity-monitoring library that supports decisions on treatment-solution design and regulation. In this project, we work with dam authorities across three states of Punjab, Gujarat and Maharashtra to build a framework for flood reduction through better reservoir control strategies. We're also developing SOTA neural surrogates for hydrological and hydrodynamic simulations.
Self-supervised learning for agricultural landscapes, uncovering the potential of remote-sensing modalities for downstream agricultural tasks.
An AI-native IDE for urban planners, enabling natural-language access to GIS and spatial analytics. We're also building MCP servers to resources such as CoreStack, GEE, IUDX and more to enable a hassle free data access, specially in India.
Past projects
OpenGeoDemand — A library to simulate quick-commerce orders and estimate optimal dark-store locations. Watch video ↗
CityCool — Methods and dashboards for tracking cool roofs, trees, and changes in lakes over time using satellite imagery.
Support & collaboration
Join
Build spatial data systems, training pipelines, evaluation tooling, and interfaces that survive contact with real deployments.
Work on multimodal learning, active learning, planning systems, and geospatial representation learning with strong downstream relevance.
Bring urban planning, climate, agriculture, hydrology, or accessibility expertise into projects that need sharper problem definitions.
Start A Conversation
We are a research lab at Plaksha University, led by Anupam Sobti. There may be a good fit where difficult spatial problems call for patient, applied research that can move from paper to practice. If you are considering funding a project, collaborating, or applying to the lab, we would be grateful to hear from you.
