Spatiotemporal Intelligence — In Development

Ask the physical world a question. Get an answer you can act on.

>

SceneQuery is the conversational layer between people and spatiotemporal data — satellite, drone, sensor, and video — turning a plain-language question into a visualized, evidence-backed answer.

Most spatiotemporal data sits behind dashboards, GIS tools, and query languages built for specialists.

SceneQuery is built on a different premise: the person who needs the answer — a grower, an analyst, a duty officer — should be able to ask for it the way they'd ask a colleague, and get back something they can actually use, not a table they have to interpret. One reusable engine, ingest → spatiotemporal index → vision-language feature extraction → natural-language query layer → visualized output, pointed at different sectors and datasets.

What's in progress

Updated weekly — see the GitHub log
In Development

Investigative Video Search

Natural-language query over a body of footage, returning timestamped, visualized results.

Queued

Infrastructure & Asset Monitoring

Fusing sensor and imagery data over time to answer plain-language risk questions.

Queued

Environmental & Resource Forecasting

Satellite and time-series data, combined to answer forward-looking questions with a visual output.

Queued

Multilingual VLM

Addressing the multilingual gap in vision-language models — regional context, low-resource-language output.

Follow the build

Code & weekly notes GitHub Runnable demos Hugging Face Build walkthroughs Video