Field notes on land, maps and AI.
How we build models for Bangladesh's land records, maps and imagery — what works, what does not yet, and what it means for fair land valuation.
Bangladesh's Proposed Land Valuation Rules, Explained
What Bangladesh proposed in 2025 for setting the minimum value of land for registration, which tax changes took effect, and the plot-level data the new method would need.
Read the post →More posts
RSS feed →
AI That Stays in the Land Office
Why our AI for land records runs offline in the office, leaves approval to people, speaks Bengali first and keeps every record's history, and what that means for procurement.
Read the post →
A Field Guide to Bangladesh's Land Records
What a khatian is, how mouza, JL and dag numbers link it to the map, why CS, SA and RS records still matter, and where mutation and online services stand.
Read the post →
Checking the Land Behind a Loan
How banks in Bangladesh check land offered as collateral, where those checks break down, and what connected khatians, mouza maps and aerial photos make possible.
Read the post →
Three Questions Behind Every Fair Land Value
Who owns a plot, where exactly it is and what is on it today: where each answer lives in Bangladesh, why it is hard to get and how our models help.
Read the post →
How We Test Our Models, and Where General-Purpose AI Falls Short
Every score in this post was measured on mouzas, map sheets and areas our models never saw in development, with the weak cases shown and general-purpose AI as a baseline.
Read the post →
The Khatian Platform, Where Records, Maps and Ground Meet
The Khatian Platform brings our three models together in an offline desktop application that takes a land office from scanned pages to published, searchable records, with people in control.
Read the post →
What the Land Looks Like Today
Using 10 cm aerial photos from 2025, our model outlines the fields, ponds and homesteads on the ground today, showing where land has changed since the old survey.
Read the post →
Turning Old Mouza Maps into Digital Parcels
Our model turns scanned mouza sheets into digital plots with their numbers, finding 95 percent of plots on 22 sheets it had never seen.
Read the post →
Reading Bangladesh's Handwritten Land Records
How we built a model that reads Bangladesh's handwritten khatians, scoring 91 out of 100 on the index fields where the best general-purpose AI we tested scored 16.
Read the post →Talk to us about land valuation.
See how our models and the Khatian Platform put the record, the map and today's land for each plot on one screen, running on a computer in your own office.