Applied AI · official statistics
Sana
Making official statistical microdata accessible through natural-language questions.
Sana is a natural-language assistant for querying official statistical microdata. At STAT.DATATHON-2026, Team Sana placed third in the AI for Official Statistics track.
01 / THE QUESTION
How can people ask questions of official statistics while respecting the privacy of the underlying data?
STAT.DATATHON-2026 was organized by the Bureau of National Statistics of Kazakhstan, with support from the CIS Interstate Statistical Committee and partners El Umiti Corporate Foundation and Alem.ai Foundation.
- 01Natural-language question
- 02Statistical microdata
- 03Answer evaluation
A conceptual view of the task and its evaluation; not a product screenshot.
02 / THE WORK
My contribution
Led development of the core AI system.
Designed the research methodology and evaluation framework.
Coordinated the team’s technical work.
Methods & focus
- Natural-language data access
- Reference-question evaluation
- Privacy-challenge testing
03 / EVIDENCE & LIMITS
What the work supports
The team’s internal evaluation reported correct answers on 60 of 60 reference questions and successful handling of 25 of 25 privacy-challenge prompts.