In 2020 Sanjay Mandava flew from Chicago to Chennai to settle a family land dispute. He spent months moving between government offices collecting a dozen records to prove the property belonged to his family.
The original sale deed was not enough. He was asked for building certificates, property tax receipts, land survey maps and more.
He went away and built Landeed, a platform that aggregates fragmented land records, verifies titles and runs legal checks across 24 Indian states.
Why the problem exists at all
This is the part that explains everything else, and it is a legal difference rather than a technical one.
"In the US or UK, when you purchase a property, the government verifies ownership," Mandava told the BBC. "In India by contrast, we follow presumptive titling — the buyer is responsible for establishing ownership. A sale deed merely records the transaction."
The consequences are enormous. Property disputes account for the bulk of civil litigation in a country where some 80% of household wealth sits in real estate.

What the company actually built
Landeed mapped large parts of India using geo-spatial technology and data from state land revenue departments, registration offices, survey records and municipal databases, linking registered properties to their official identification and survey numbers.
An AI layer then reads the titles, extracts information and spots inconsistencies in minutes — work that takes lawyers and brokers several days.
It now holds 1.8 billion documents, the largest repository of title data in the country, and is used by millions of individual and enterprise customers.
The pattern this belongs to
Unlike the US and China, which dominate frontier language models, India's AI activity is happening downstream — in data centre infrastructure and consumer applications rather than research.
Bengaluru-based YouTube teacher Gaurav Sen built an app that detects potholes from a dash-cam using satellite navigation and an AI vision model, identifies the civic officials responsible, and sends them the data. At least half a dozen similar services now exist.
"A municipal corporation can take the app and use it on public buses and taxis, to access real-time, updated data on road conditions," Sen told the BBC.
In Mumbai, Fractal Analytics — one of the few listed Indian AI companies — runs a tool called Vaidya in hospitals operated by the civic authority. It gives patients simplified summaries of their lab reports in multiple languages, points them to nearby facilities, and gives doctors AI-guided pre-consultation support. Administrators get a public health dashboard built from anonymised lab data, which shows emerging infections and disease trends.
The money, and the catch
Funding for Indian AI start-ups is up 265% year on year according to Inc42, with more than $1bn raised in the first nine months of this year — in an otherwise muted environment.
Commercialising remains hard, and the reason is one every Bangladeshi software company will recognise.
"The cost of building in India is sometimes much higher than what Indian customers are willing to pay," Fractal co-founder Srikanth Velamakanni told the BBC. The computing power to run a model costs the same whether the company is in India or the US — but American customers pay more.
Why this matters here
The structural position is nearly identical: fragmented multilingual data, cost-sensitive customers, real administrative problems that nobody has digitised, and no realistic path to competing on frontier research.
India's answer was to stop trying to compete there and go after problems that only a local company understands well enough to solve. Land records, potholes, lab reports in the patient's own language.
That is a strategy available to Bangladesh too, and it does not require a single new model to be trained.




