A research toy / modelled after karpathy.ai/jobs
This tool visualises roughly 70 million working people across 55 occupations and 14 sectors of the Bangladesh economy, built from the Bangladesh Bureau of Statistics Labour Force Survey 2024 plus sector sources. Each rectangle's area is proportional to how many people do that job. Colour shows the metric you pick below: sector, monthly pay, formal versus informal status, digital AI exposure, or automation and robotics exposure. Hover any tile for the detail and the reasoning.
The headline finding is the mirror image of the United States. In the US economy a huge share of work is digital knowledge work, so the famous Karpathy treemap glows red. Bangladesh is dominated by farming, garment stitching, driving, trading and domestic work, so on the digital AI layer it stays mostly green. The sharper risk here lives on the automation layer, where the readymade garment sector alone holds about 4 million jobs.
Pay, formal status and sector come from survey and sector data. The two exposure layers are scored 0 to 10 by the same logic Karpathy used: how much will the technology reshape this job, counting both direct automation and indirect productivity effects.
Digital AI exposure is high when the work product is fundamentally digital and can be done on a computer. Automation and robotics exposure is high when the work is repetitive physical labour that a machine or robot could take over, which is the layer that matters most for Bangladesh's factories and fields.
A high score does not mean a job vanishes. Garment work scores high on automation, yet sewing remains hard to fully roboticise and demand could still grow. The scores ignore demand elasticity, cost of capital in a low wage economy, regulation, and social preference for human workers, all of which run strongly in favour of Bangladeshi labour for now.
Employment counts are estimates built from LFS 2024 sector shares and sector sources, not an official occupation by occupation table, which the BBS does not publish at this resolution.