Zindi: Barbados Historic Handwriting AI

Zindi is an African data science platform, and this challenge is run for GovTech Barbados, the state-owned digital agency. Barbados has digitised thousands of colonial-era handwritten deeds, wills and census records, and wants a machine learning model that turns scanned pages of faded, irregular handwriting into machine-readable text. Scoring is a 50/50 blend of word error rate and character error rate.

Entry is open to all nationalities, so Cameroonian students compete on equal footing. You may work alone or in a team of up to four.

Read the prize split carefully. The pool is USD 25,000, but USD 5,000 of it is reserved for the highest-placed Barbadian, so non-Barbadians are effectively competing for USD 20,000 across three podium places.

Two practical catches. Only open-source tools are allowed — no paid services and no free trials that need a credit card. And if you finish in the top ten you have 48 hours to submit reproducible code; failing to respond means disqualification.

Nothing is asked of you at entry, though winners must later show ID and bank details to be paid. Zindi publishes the close date as 5 October 2026 without a clock time, so do not leave your final submission to the last hours.