AI & ML interests

None defined yet.

Recent Activity

MosesJoshuaCoker  updated a Space 5 days ago
NovaxAILab/README
MosesJoshuaCoker  published a Space about 1 month ago
NovaxAILab/README
View all activity

Organization Card

Novax AI Lab

Building AI that understands Africa's languages.

Speech Recognition • Text-to-Speech • Translation • Language Models • Open Source


"Every language deserves a place in the future of AI."

Website · Krio AI · Models · Research

🌍 About Us

Novax AI Lab is an AI research and product company based in Sierra Leone.

Our mission is to make artificial intelligence work for every African by building technology that understands, speaks, and translates indigenous languages — starting with Sierra Leonean Krio and expanding across the continent.

The website is the canonical documentation layer for everything below: novaxailab.com.


✅ What's Available Today

🎙️ Krio Speech-to-Text

A dedicated speech recognition model for Krio, fine-tuned from openai/whisper-small on ~30 hours of Krio speech. WER 5.05% / CER 2.18% on clean read speech. Apache-2.0. → huggingface.co/MosesJoshuaCoker/best_for_now

🔊 Krio Text-to-Speech — beta

Neural TTS that turns written Krio into natural spoken audio.

🧩 Phos — speech-to-text with translation

Transcribes and translates Krio and English in a single workflow. → novaxailab.com/models/phos

Feedback from real usage is shaping the next releases.


🚧 What We're Building Next

More languages for speech recognition & TTS

  • 🇸🇱 Sierra Leone: Mende, Temne, Fula, Susu, Limba
  • 🇱🇷 Liberia: Kpelle, Kissi
  • 🇬🇭 Ghana: Akan (Twi, Fante, Akuapem), Ewe, Dagbani

🌐 Machine Translation

  • English ↔ Krio
  • Krio ↔ Mende, Krio ↔ Temne, Mende ↔ Temne
  • English ↔ other African languages listed above

🎧 Speech-to-Speech Translation

Speak naturally, listen naturally — no text step in between.

🤖 AI Assistants

Applications planned for education, healthcare, government, customer support, agriculture, and finance.


🔬 Research Areas

  • Speech recognition & text-to-speech for low-resource languages
  • Machine translation & speech-to-speech translation
  • Multilingual NLP and language models
  • Speech datasets and evaluation for African languages
  • Voice agents and applied generative AI

Research write-ups: novaxailab.com/research


🌍 Why We Exist

Hundreds of millions of people speak languages that modern AI barely understands. We're changing that — making African languages first-class citizens in the age of AI, with open models and measurable results.


🛠️ Open Source

We believe open collaboration accelerates innovation. As models and datasets move out of beta, this organization will host:

  • Models
  • Datasets
  • Training pipelines
  • Research projects
  • Developer tools & APIs
  • Sample applications

Contributions from researchers, developers, linguists, and students are welcome.


🤝 Collaborate With Us

We're looking to work with AI researchers, ML engineers, linguists, universities, NGOs, governments, startups, and open-source contributors.


📫 Connect

🌐 Website: novaxailab.com 🤗 Hugging Face: huggingface.co/NovaxAILab 💻 GitHub: github.com/Novax-AI-Lab ✉️ Email: [email protected]


Building AI for Every African Language.

Made with ❤️ in Sierra Leone 🇸🇱

models 0

None public yet

datasets 0

None public yet