Ask Rwanda's data.
Get the proof.

AI agent for official NISR statistics

Ask in English or Kinyarwanda. Baza finds the right NISR table, runs the query, and checks every number before you see it.

The problem

The numbers exist.
Finding them is the hard part.

01

Scattered

NISR data sits in 7+ portals and hundreds of Excel and PDF tables. NISR's own strategy lists the lack of an integrated repository as a weakness.

02

Mostly in English

District workshops reported that only some publications are in Kinyarwanda, and that officials struggle to navigate NISR's channels.

03

Chatbots guess

General AI tools give confident but wrong figures. Official statistics need an answer that shows its source and its work.

How it works

Ask a question.
Get the evidence.

It plans
The agent picks the right tables from 35 cleaned NISR tables and plans its queries.
It queries
Read-only SQL runs in parallel across the Census, EICV7 and the Labour Force Survey, joined by district.
It proves
Every number in the answer is checked against the results and tagged with its NISR table. If the data cannot answer, it says so and points to the right NISR survey.

Sample answer Verified

Nyamagabe has the highest poverty rate, 51.4%, against 27.4% nationally (EICV7 2023/24).

Source: EICV7 2023/24, Table A9.23, NISR

AGENT · INVESTIGATION

District brief

Pick a district. The agent plans what to check, compares the district with its province and Rwanda across the Census and EICV7, and writes a one-page brief for Imihigo planning, community meetings or radio. Every number is checked.

FEEDBACK LOOP FOR NISR

Data-gap radar

Every question the published tables could not answer becomes a signal for NISR: what people need, and which NISR survey the agent pointed them to. No names or IP addresses are stored.

UNANSWERED QUESTIONS · MOST RECENT

NISR SOURCES PEOPLE WERE SENT TO

MOST USED TABLES

ONE DATABASE · 4 NISR PUBLICATIONS

Data catalogue

Every table the agent can query, cleaned from NISR's Excel files and joined on a common district key. CC BY 4.0, National Institute of Statistics of Rwanda.

ABOUT

Why Baza NISR

Baza means "ask" in Kinyarwanda. NISR publishes rich district statistics, but they are spread across separate portals and hundreds of Excel and PDF tables, mostly in English. NISR's own strategy (NSDS4, 2024 to 2029) lists the absence of an integrated data repository and too little use of Kinyarwanda as weaknesses, and calls for AI-assisted dissemination. Baza NISR is a working version of that layer.

How the agent works

It pre-selects likely tables from a catalogue of 35 cleaned NISR tables, then plans its steps.
It writes read-only SQL, runs independent queries in parallel, and joins surveys on the district key.
When the tables cannot answer, it searches NISR's own microdata catalogue and points to the right survey.
A verifier checks every number in the answer against the query results. Unmatched numbers are sent back once for correction, then shown as a warning.
Answers can be switched to Kinyarwanda at any time. A translation is shown only if it keeps exactly the same numbers.
Unanswered questions feed the data-gap radar for NISR.

Data

5th Population and Housing Census (RPHC5), 2022, main indicator tables.
EICV7 2023/24 main indicators and the Utilities and Amenities thematic tables.
Labour Force Survey quarterly tables up to Q2 2026.
District boundaries: Open Data Rwanda (NISR) via geoBoundaries, CC BY 4.0.

Source: National Institute of Statistics of Rwanda, statistics.gov.rw, CC BY 4.0.

Limitations

Census 2022 and EICV7 2023/24 differ in year and method, so comparisons across them are indicative.
EICV7 district estimates come from a sample and have confidence intervals.
Only loaded tables can be answered with numbers. Kinyarwanda answers are machine translated.

Embed on any NISR page

One line adds an "Ask NISR" button that opens the agent in a panel:

<script src="/embed.js" defer></script>

Team

Abdulmujeeb Tiamiyu and Carlos Ngoga, Carnegie Mellon University Africa. NISR 2026 Big Data Hackathon, Track 3: Open Innovation.

Ask Rwanda's data.Get the proof.

WHERE NISR HAS THIS

DATA

SOURCES

The numbers exist.Finding them is the hard part.