AI built for humanitarian and development work.
Halal Numeriq builds AI applications, and trains and fine-tunes models, for the organizations serving communities in Northern Nigeria and beyond.
Based in Sokoto, Nigeria
The problem
The work is urgent. The data and tools lag behind.
Field data arrives messy.
Errors are caught weeks later, after enumerators have left.
Critical knowledge is buried in documents.
Budgets, plans and response SOPs run to hundreds of pages, and when a decision or an emergency comes, nobody can find the answer fast.
AI doesn't speak the community's language.
Hausa and Fulfulde voices are often left out of the tools.
What we build
Four tools for the people doing the work.
All four are in development. We want to shape them with the organizations that will use them, so we are open to pilot partners.
- In development
Survey QA System
Survey errors surface weeks late : flag them while teams are still in the field.
Flags inconsistent ODK/Kobo submissions, outliers and suspicious enumerator patterns in near-real time, and explains each issue in plain language so supervisors can fix it while teams are still in the field.
Talk to us about piloting: Survey QA System - In development
Budget Intelligence (RAG)
Budget answers hide in long documents : ask in plain language and get cited answers.
Ask questions in plain language across budget documents, workplans and expenditure reports, and get answers that cite the exact source page. Built for planning teams, development partners and budget-transparency work.
Talk to us about piloting: Budget Intelligence (RAG) - In development
Emergency Response Assistant (RAG)
In a crisis, nobody has time to search SOPs : get fast, cited answers instead.
Fast, cited answers from SOPs, contingency plans, situation reports and assessment data, so responders can act quickly during floods, displacement and outbreaks.
Talk to us about piloting: Emergency Response Assistant (RAG) - In development
Hausa Language Models
General AI handles Hausa poorly : fine-tune open models that work in it.
Fine-tuning open models for Hausa, covering transcription, translation and classification of community feedback, so AI tools work for the people they're meant to serve.
Talk to us about piloting: Hausa Language Models
Train & fine-tune
We fine-tune where general models fall short.
Our first focus is Hausa: speech-to-text, translation and text classification for community feedback. Every model is evaluated against real field data before use.
Now
Hausa
Speech-to-text, translation and text classification for community feedback.
Next
Fulfulde
Extending the same language work to Fulfulde.
Next
Sector-specific classifiers
Classifiers trained for humanitarian data.
Evaluated on real field data
Before any model is used, we test it against real data from the field, not only public benchmarks.
Retrieval-augmented generation (RAG)
We ground AI answers in an organization's own documents, with citations, so every answer can be checked against its source.
Responsible AI
Guardrails come first, not last.
Data protection by design
Built to comply with the Nigeria Data Protection Act 2023, with anonymization before data reaches any model, data minimization, and clear data agreements.
People decide
AI flags and drafts, and humans review, especially for anything affecting aid.
Do no harm
We assess risks to communities before deployment.
Transparency
We explain what our tools do and where they can fail.
Who we work with
Built for the organizations serving communities.
- NGOs and INGOs
- UN agencies
- Research and survey firms
- Donors and foundations
- Development programs
Why HNQ
Rooted in the field.
HNQ is rooted in field experience with household surveys, M&E and humanitarian data systems in Northern Nigeria. We build for enumerators, M&E officers, program managers and the communities they serve.
Bilyaminu Bawan Allah
Founder & CEO
Computer scientist and humanitarian data practitioner from Sokoto, building AI that works for the people and organizations serving Northern Nigeria.
Partner with us
Let's build AI that serves communities.
Tell us about your program, your data and the problem you want to solve.