Surely read this.
Here is my extremely unpopular opinion/statement:
Blockchain is overrated.
No use case suggested so far actually requires a distributed secure data store, and hence doesn’t need blockchain.
All hype is killing what we really require. https://twitter.com/rmantri/status/1352627068618616839

This is the basics in IT solution:
Don’t use a hammer for everything just because everyone is talking about hammer.

Don’t use a wrench where you need a screwdriver.

If you can reach the ripe mango, pluck it with your hand.

Let’s take land records as an example:
What do we really lack:
We do not exactly know who owns a land. (Prof @bibekdebroy had written about it)

There are a lot of civil suits on land, which are left pending because litigation costs more than eventual share.

What we really want:
Enforce right to property.

We need every piece of land mapped to an aadhar number/pan/voter id.
We need geo-tagging of these properties.

Over the years, the area on the sales deed(the only authentic proof of land ownership now) doesn’t really match with land in possession in many places.
There is no clear idea what is public land and what is government owner land.

Let’s assume we figure out a way to get the data.
Now, the next question to ask is: what’s the data used for.

A) giving property taxes.
B) establishing ownership
C) for some loan or similar financial purposes
D) transfer of ownership (either gift, sale or inheritance).

Of the four things that we can do with land data, only transfer requires writing on the digital land record.
You might say hypothecation too may require a write(not necessary, but I’ll accept that claim).
All others are read only work.

Now, let’s come to the transfer part.
Every such transfer has to be registered in the sub-registrar office. Let’s say we managed to ensure that the registration(basically rewriting the land data is done with authentication, verification and validation).

This transfer is of two kinds(actually three).
i) Transferring the whole plot.
ii) Transferring part of plot.
iii) (related to ii) merging two neighboring plots(or expanding a plot by buying neighbors plot).

None of this requires blockchain.
Just because everyone talks about blockchain doesn’t mean that blockchain is applicable here.
(But, trust me, if i have to sell a digital land system to sell, i will use the name of blockchain to sell it even if I don’t use it).

I will accept that a distributed file system makes sense.
Frankly, a BigTable is more than enough for this. Even that is an overkill. We would love distributed data with a lot of replication purely from a reliability perspective. Nothing else.
Even a mysql database is good jugad.

Basically, we need to choose the tool based on what we truly need.
I suggested an unstructured/semi-structured data store primarily because we can then choose to add satellite imagery.
Yes. Standard GIS solutions already exist.

That brings us to my last point.
Listen, our major problem is just validating ownership and adding them in anything digital. Even a simple spreadsheet is good enough.
First task to do is this.
It will also help resolve a lot of land disputes in the courts.

You can throw almost any usecase from a government perspective and i can use the same technique to tell you why we don’t want either block chain or AI(that’s a whole different level of abuse) and it’s current cousin ML.
Cloud we will end up using. It helps abstract a lot of stuff