I built this because I'm tired of ML tools that don't work for us.
Tired of tokenizers that butcher our languages. Tools that assume unlimited GPU. Models that don't know our names.
Nigeria deserves ML infrastructure built by us, for us.
More coming → Spread the Word Nigeria!
NaijaML v0.1.1 is live.
pip install naijaml
GitHub: https://t.co/Y8bn7UTaIQ
HuggingFace: https://t.co/J0fXMbdXUy
PyPI: https://t.co/S9G21RHnAA
Use it. Break it. Contribute.
Africa's 4th Industrial Revolution can't run on borrowed infrastructure.
Africa spends $55B/yr on security. Every intelligent system — drones, fraud detection, diagnostics — runs on ML.
But the ML infrastructure wasn't built for Nigerian data or languages.
Sovereign AI needs sovereign ML infrastructure. NaijaML is that foundation layer.
Design philosophy:
Every decision starts with "Does this work on a laptop in Lagos with 4GB RAM and intermittent power?"
→ CPU-first, GPU optional
→ Offline after first download
→ 17MB total models
→ Deps: numpy, requests, tqdm
→ Honest accuracy numbers
Sentiment analysis across Nigerian languages:
analyze_sentiment("This film too sweet!") → positive
analyze_sentiment("Wannan fim din yana da kyau") → positive (Hausa)
72% accuracy — honest number. 4.3MB. Transformer upgrade coming.
Most requested tool in Nigerian NLP:
from naijaml.nlp import diacritize_dot_below
diacritize_dot_below("Ojo lo si oja") → 'Ọjọ lo si ọja'
97.5% accuracy. 6.4MB. No GPU. Works offline.
Nigeria has 237M+ people. 500+ languages. Largest tech ecosystem in Africa.
But standard ML tools:
→ Destroy Yorùbá diacritics
→ Don't recognize Nigerian names
→ Think Pidgin is English
→ Strip "sha" and "sef" as typos
BTW, Pidgin has 120M+ speakers.
Standard NLP tools can't tell "day" from "rain" in Yorùbá. They strip diacritics. They think Pidgin is broken English.
So I built NaijaML — open-source ML tools for Nigerian languages.
Live on PyPI today: pip install naijaml
🧵
The puzzle consists of 5 different-colored houses in a row, each lived in by a resident of a different nationality. Each resident owns a different pet, prefers a different drink, and smokes a different brand of cigarettes than the others. Who owns the zebra, who drinks water?
Apparently, Albert Einstein left an assignment for the world before he left. Only 2% of the World Population is rumored to be able to solve the puzzle without Ai.
can you solve it?