V
Loading...
Loading...
Loading...
Our research is published in top-tier venues and made available to the broader research community. Explore our papers, reports, and whitepapers advancing AI for African impact.
45+
Published Papers
6
Research Areas
15+
Top-Tier Venues
10+
Best Paper Awards
Adeyemi, G., Asante, K., Okonkwo, A., Banda, M., et al.
ACL 2024 (Main Conference)
We present AfroLM, a large language model trained on 30 African languages using a novel multilingual pre-training approach. AfroLM achieves state-of-the-art results on multiple African language benchmarks, improving upon existing multilingual models by 15-25% on tasks including named entity recognition, sentiment analysis, and machine translation.
Mwangi, J., Okonkwo, A., Hassan, F., Kimani, D., et al.
NeurIPS 2024
We conduct the first large-scale study of fairness in machine learning-based credit scoring across five African countries. Our analysis reveals significant disparities in model performance across demographic groups and proposes novel fairness constraints that improve equity while maintaining predictive accuracy.
Kimani, D., Mutua, L., Banda, M., et al.
MLSys 2024
We present a comprehensive framework for deploying machine learning systems in environments with intermittent connectivity and resource-constrained devices. Our approach achieves 10x model compression while maintaining 95% accuracy and enables fully offline operation with intelligent synchronization.
Adeyemi, G., Hassan, F., Wanjiku, S., et al.
CHIL 2024 (Conference on Health, Inference, and Learning)
We present clinical validation results for a Swahili voice-based symptom checking system deployed across 200 rural health facilities in Tanzania. The system achieves 87% accuracy in triage classification compared to physician assessments, while reducing average triage time by 75%.
VE.KE Research Team
VE.KE Publications
Our annual report on the state of artificial intelligence adoption and development across Africa. This year's report covers AI policy developments in 15 countries, investment trends, key use cases, and emerging opportunities for AI-driven transformation across sectors.
Adeyemi, G., Okonkwo, A., Banda, M., et al.
AfricaNLP Workshop @ ICLR 2023 (Best Paper Award)
We present SwahiliBERT, a BERT-based language model trained specifically for Swahili. Using a novel data collection approach that combines web crawling with community partnerships, we achieve 23% improvement over multilingual baselines on Swahili NLU tasks.
Banda, M., Adeyemi, G., Kimani, D., et al.
EMNLP 2023
Code-switching between African languages and English is pervasive in social media. We present a novel approach to handling code-switched text that improves NER performance by 31% and sentiment analysis by 24% on a new benchmark dataset of East African social media.
Kimani, D., Mwangi, J., Okonkwo, A., et al.
KDD 2023
We present a graph neural network approach to detecting fraud in mobile money systems. By modeling transaction networks and user relationships, our system identifies fraud rings with 85% precision while reducing false positives by 50% compared to rule-based systems.
Okafor, Z., Banda, M., Mutua, L., et al.
CVPR Workshop on AI for Agriculture
Using satellite imagery and multi-spectral analysis, we develop a system that detects crop diseases an average of 3 weeks before visible symptoms. Deployed across 500,000 hectares in Kenya, the system has helped 100,000+ farmers reduce crop losses by 40%.
Asante, K., Adeyemi, G., Okonkwo, A., Hassan, F.
VE.KE Whitepaper
We present a comprehensive framework for responsible AI development and deployment in African contexts. The framework addresses fairness across diverse populations, transparency requirements, data privacy considerations, and community engagement practices.
Banda, M., Okafor, Z., Mutua, L., et al.
Remote Sensing of Environment
We develop a satellite-based crop yield prediction system tailored for the fragmented landscapes of smallholder farms in East Africa. Our approach achieves 82% accuracy in yield prediction, enabling better planning for farmers and agricultural organizations.
Banda, M., Kimani, D., Adeyemi, G., et al.
EMNLP 2023
We present techniques for deploying NLP models on entry-level smartphones common in African markets. Through knowledge distillation and quantization, we achieve 10x compression while maintaining 95% of original model accuracy, enabling on-device inference in under 50ms.
Hassan, F., Okonkwo, A., Wanjiku, S., et al.
Lancet Digital Health
We present results from a multi-site validation of AI-assisted TB detection across 50 health facilities in Kenya and Tanzania. The system achieves 85% sensitivity and 90% specificity, comparable to expert radiologists, and has been integrated into national screening programs.
Mwangi, J., Kimani, D., Asante, K., et al.
ACM FAccT 2022
We develop privacy-preserving machine learning techniques for credit scoring using mobile phone data. Our federated learning approach enables model training without centralizing sensitive data, while achieving credit prediction accuracy comparable to centralized models.
Adeyemi, G., Asante, K., et al.
LREC 2022
We describe our efforts to build NLP datasets and resources for 15 low-resource African languages. Through community partnerships and novel annotation approaches, we create benchmark datasets that are now publicly available to accelerate African language NLP research.
Our publications span multiple research areas, each focused on solving real problems for African communities and businesses.
We're always looking for research partners and collaborators to advance AI for African impact. Get in touch to explore opportunities.