SciPy India · three talks · every demo runs on real, live Indian data

The mathematics hiding under maps

One idea, attacked from three directions: maps are not special software. They are counting, measuring, and connecting dots, done carefully. Each talk has working demos you can run yourself, built on a satellite photo of the Brahmaputra, 38 million real buildings, live fire alerts, Delhi's air sensors, and the shape of the Himalaya.

Talk 01 · educational

Under the Map: Rebuilding GIS Algorithms with SciPy

Six things everyone treats as expensive mapping-software buttons, rebuilt in a few lines of Python each. Finding water in a satellite photo, distance zones, the nearest village to every fire, mountain paths, a pollution map of Delhi, and where the rain goes.

6 demosndimage · spatial · sparse · interpolate read →
Talk 02 · engineering

When GeoPandas Stops Scaling

One innocent question: find the nearest road for every building in Bihar. 38 million buildings, 11 million road points. The obvious approach needs 69 days and a 3,400-terabyte table. The right representation answers it in 5 seconds on one machine. This talk is about the difference.

4 demos + a data war storycKDTree · sparse · csgraph · memory read →
Talk 03 · conceptual

The Map Is a Matrix

Put a map on screen. Take it apart until only mathematics is left: terrain becomes a grid of numbers, towns become a cloud of points, roads become a graph, and "where should the next hospital go?" becomes an equation with an exact answer. Then bring the map back to look at what you solved.

4 demosarrays · Voronoi · graphs · optimization read →

Talk 04 · Field stories

One real Indian problem each, solved live from raw data to a decision. Less lecture, more lab notebook: every number below came out of an actual run.

Story A · the flagship

One Flood: From Raw Pixels to a Relief Plan

The September 2025 Kosi flood in four acts: radar finds 270 km² of water through the clouds, 25,428 buildings turn out to be in it, 195,565 lose their road access, and an equation places the relief camps.

4 actsread →
Story B · land records

Whose Land Flooded?

The compensation question, answered from 11.2 million raw cadastral parcels: 2,945 flooded, village by village. Plus the uncomfortable measurement: the cadastral map itself sits 20–34 metres west of reality.

2 partsread →
Story C · urban expansion

Which Way Is Patna Growing?

Two mosaics, eight years apart: 37.5 km² of new construction marching east and west along the Ganga, found by a rule you can read — including the texture trick that stops sandbars posing as concrete.

1 sittingread →
Story D · accessibility

How Far Is Help?

Every building in Bihar, stamped with its true road distance to the nearest of 1,276 hospitals, in one 1.4-second sweep. 7.4% are beyond 20 km, and the single best new hospital site turns out to be Bagaha.

1 sittingread →