After the Bhashini posts, a reader asked for the same treatment of PM GatiShakti. It is a very different kind of platform: not AI models, but one shared map for planning India's infrastructure. A simple, fact-checked explainer.
After I wrote about how Bhashini works, a reader asked me to do the same for PM GatiShakti. Happy to, but with one important clarification up front, because the two are very different animals.
Bhashini is a set of AI models for language. GatiShakti is not an AI model at all. At its heart it is one giant shared map that every infrastructure ministry can see and draw on. Both are national digital platforms, and both share the same good instinct of building something public that everyone can stand on. But the technology underneath is not the same, and it would be wrong to treat them as if it were.
As with the Bhashini posts, everything here is from public sources, listed at the bottom so you can check it yourself.
PM GatiShakti is, in plain terms, one shared map of the country where every ministry can see everyone else’s infrastructure, both what exists and what is planned, so they stop planning in the dark and stop tripping over each other.
That is the whole point, and to see why it matters, you have to picture the problem it was built to fix.
For decades, each part of government planned its own infrastructure on its own. The roads ministry planned roads. The railways planned tracks. Separate teams handled ports, power lines, gas pipelines, and telecom. Each had its own maps and its own plans, and almost none of them could easily see the others.
The result is the thing every Indian has watched happen on their own street. A road gets laid, and a month later it is dug up again to put down a pipe that another department had always planned to lay. Multiply that across a country and it becomes enormous waste, in money and in time.
GatiShakti exists to end that blindness. Put every department’s existing assets and future plans onto one common map, and the clashes become visible before anyone breaks ground, not after.
Now the evidenced facts.
PM GatiShakti National Master Plan was launched on 13 October 2021. It was developed by BiSAG-N, the Bhaskaracharya National Institute for Space Applications and Geoinformatics. It is overseen by the DPIIT, the Department for Promotion of Industry and Internal Trade, whose Logistics Division acts as the secretariat and technical support unit.
On the technology side, the platform is built using open-source technologies and hosted on MeghRaj, the government of India’s cloud. So it is a government-run system, running on government cloud, built by a government institute.
The platform is built on GIS, which stands for Geographic Information System. The simplest way to understand GIS is to think of a smart map where every kind of information sits on its own transparent layer that you can switch on or off.
One layer is roads. Another is railway lines. Another is rivers, another is forests, another is power transmission lines, another is land records, and so on. On their own each layer is just one slice of reality. Stacked on top of one another on the same map, they suddenly show you the full picture of a place: what is there, what is protected, what is already planned, and what would clash with what.
A few concrete details about those layers, from the public record. The platform is commonly described as having more than 200 planning and analytical layers, and reports describe over 1,400 data layers integrated across central and state governments. Each state has mapped on the order of 29 essential layers covering things like land records, forests, wildlife and eco-sensitive zones, water resources, flood maps, power lines, mining areas, and roads.
Two more pieces make it work in practice.
First, satellite imagery from ISRO is fed in, which gives a real, current view of the ground and lets the system track the actual progress of projects over time, rather than relying on someone’s status report.
Second, each ministry logs in and maintains its own layers. Departments update their own data regularly, and because everyone is looking at the same map, every department can now see, review, and monitor the others. The schemes already folded in include large ones like Bharatmala, Sagarmala, inland waterways, and UDAN.
So when a new project is being planned, the planner can see in one place that the proposed road would cut through a reserve forest, or run alongside a planned pipeline, and adjust before a rupee is spent. The public benefit cited again and again is exactly this: fewer cost overruns and delays, because alignment conflicts are caught early using the geospatial data.
With Bhashini I could point to a specific base model, IndicTrans2, and tell you how it was trained. People naturally ask the same of GatiShakti, and here the honest answer is different.
GatiShakti does not have a base model in that sense, because it is not a trained AI model. Its foundation is the GIS platform, the satellite imagery, and the shared data layers, not a neural network. There is geospatial analysis and some AI used for monitoring and decision support, but the core of the thing is the map and the discipline of putting everyone’s data on it in a common form. The intelligence is in the integration, not in a single model.
This is the natural follow-up, since I argued earlier that a department can self-host the open Bhashini models on its own GPU. The honest answer for GatiShakti is no, not in the same way.
Bhashini gives you open model files you can download and run anywhere. GatiShakti is a government platform with ministry logins, hosted on MeghRaj, not a model you pull down and serve on your own box. It is reported to use open-source GIS technology underneath, and a public version of the platform has been launched to open up access to some of the geospatial and infrastructure data. But you do not fork GatiShakti the way you fork an open model. It is shared infrastructure you connect to and contribute to, not software you take home and run.
That difference is worth understanding, because it changes what “using” each platform even means.
Strip away the technology and the two platforms are chasing the same idea from different directions. Bhashini removes the language barrier between a citizen and a service. GatiShakti removes the coordination barrier between one arm of government and another.
One is built from AI models you can download. The other is built from a shared map you all agree to keep honest. Different engineering entirely, same instinct: build a common public foundation so that everyone, citizens and departments alike, stops working blind.
For a country the size of India, getting every infrastructure ministry to plan on one map is not a small technical trick. It is a genuinely hard act of coordination, and the map is what makes it possible.