AI Research

Our research spans multilingual AI, ethical AI, healthcare systems, and domain-specific automation — with publications and open contributions to the African NLP community.

Active Research Areas

  • Multilingual AI & African Language Models
  • Healthcare AI & Biomedical Devices
  • AI-Powered Business Automation
  • Ethical AI & Bias Mitigation

Research Contributions

  • PEPLER with reparameterization for explainable recommendations

    Integration of reparameterization to PEPLER's implementation (DIV-3.55, FCR-0.11, BLEU-4 0.8197)

  • Automatic Post-Edit (APE) Translator with Online Adaptation

    Designed and trained APE model. BLEU-4 0.43, TER 0.66.

  • Adapting Large Language Models for Collaborative Semantic Recommendations

    Modified original implementation to support unique user index generation.

  • JHU++ Image Crowd Counting (open source implementation)

    Implemented a confidence-guided deep residual crowd counting model using PyTorch.

  • Evaluating Bias in Large Language Models For African Languages

    Language bias in LLMs, comparing metrics between African, English and European texts.

Affiliations

  • Masakhane member
  • DataFest Africa
  • Igbo AI

Grants & Recognition

  • ML Collective Compute Grant — Awarded for SabiYarn-125M development
  • Guest speaker at DataFest Africa 2024 on "Transforming Patient Care through Innovation, Ethics, Challenges and Future Prospects"

Full research profile on the founder page.