Alukwe Jones Terah
Data Scientist / AI Engineer · Nairobi, Kenya
Data Scientist / AI Engineer · Nairobi, Kenya · Available for opportunities
Building AI
for a greener,
smarter world.
NLP, Generative AI, and climate-tech systems. Presenter at NeurIPS 2024 and Data Science Africa 2025. Passionate about applying machine learning to Kenya's most pressing challenges — from carbon emissions to agricultural resilience.
Skills
Experience
- Engineered document processor reducing QC by 85%
- Built paper-size classification algorithm
- Engineered an intelligent document processor that cut manual QC effort by 85%.
- Built an algorithm for automatic paper-size classification and document categorisation.
- Provided strategic AI advisory on implementation strategies and best practices.
- Deployed AI chatbot reducing response times by 60% and increasing engagement by 35%.
- Building a Generative AI model for real-time carbon emission monitoring across 500+ data points.
- Conducted analytics on agricultural datasets to surface sustainability insights.
- Produced monthly reports on public Wi-Fi utilisation trends across multiple countries.
- Monitored network health and escalated connectivity issues to minimise service disruption.
- Collected data for the Kenya Digital Economy Acceleration Project (KDEAP) across government MDAs.
- Ensured high data integrity and compliance with digital reporting standards.
- Resolved 10+ IT support tickets weekly, consistently meeting SLA requirements.
- Reduced network downtime incidents by 10% through reliability improvements.
Projects
Fine-tuned AfriBERTa for NLI and open-domain Question Answering in low-resource African languages. Training pipeline combined open-source datasets with synthetic data from Gemini, Claude, and ChatGPT.
Fine-tuned AfriBERTa for NLI and open-domain Question Answering in low-resource African languages. Training pipeline combined open-source datasets with synthetic data from Gemini, Claude, and ChatGPT.
Fine-tuned AfriBERTa for NLI and open-domain Question Answering in low-resource African languages. Training pipeline combined open-source datasets with synthetic data from Gemini, Claude, and ChatGPT.
Research contributor on an instruction-tuned model optimised for multilingual African language capabilities. Evaluated against AfriMGSM benchmarks using PyTorch and Hugging Face Transformers.
Intelligent document processor automating extraction, categorisation, and archiving of statistical data from complex government documents at DISI.
Automated support agent that resolved 80% of FAQs, cut response times by 60%, and improved client satisfaction at scale.
Generative AI + blockchain system for real-time CO₂ monitoring. Piloted with 32 industrial companies across Kenya.
Computer vision model for early identification of coffee crop pests and diseases to protect smallholder farmers.