Loading
Loading
AI/ML Specialist and Health Data Scientist experienced in machine learning, generative AI, and data analytics. Proven track record in developing predictive models, LLM-based applications, and decision-support systems for climate, health, and public-sector challenges. Published researcher and project lead with expertise in Python, NLP, time-series forecasting, and applied AI for social impact.
Developed a hybrid ML-BDN framework using XGBoost-derived prior (Spearman r = 0.926 vs 0.266 for best hand-crafted prior) for optimising contamination sensor placement in sewer networks. Preprint submitted to Elsevier.
Fine-tuned Llama-2-7b, Gemini 1.5 Pro, and GPT-4o on 2,020 IDSR instruction-response pairs for disease surveillance in resource-limited settings. RAG-augmented Llama-2 achieved 97% accuracy on priority disease case definitions within 8 GB VRAM (75% reduction in memory requirements).
Secured and led an NCST-funded research grant studying how generative AI (Llama-based LLMs) can improve equitable access to secondary education in Malawi, focusing on the agriculture subject. Designed full research methodology, managed multi-person team. Published in Wiley's EJISDC (DOI: 10.1002/isd2.70062).
Built drought early warning model for Malawi using 40 years (1980-2020) of NASA POWER and NOAA climate data. Computed Palmer Drought Severity Index (PDSI) and compared ARIMA, STL, ETS, LSTM, XGBoost. Selected XGBoost achieved R² 0.769, MAE 2.55 with regional breakdowns. Integrated into interactive Power BI dashboard.
Centre for Development & Management Consulting — Tobacco Harm Reduction Project
Leading M.E.A.L. framework design and implementation across 57 health facilities. Managing multi-source datasets (335 providers, 640 patients, 267 LMS users). Conducting longitudinal, comparative effectiveness, and logistic regression analyses.
Central Poultry Malawi
Designing data warehouses and Power BI dashboards. Applying ML models for forecasting and business performance optimisation.
The Lighthouse Trust — Kamuzu Central Hospital (HIV Programme)
Managed viral load datasets, designed KoboToolbox data collection tools for NCAP field nurses, and generated routine HIV programme reports for clinical teams.
Mzuzu University
Fine-tuned Llama-2-7b, Gemini 1.5 Pro, and GPT-4o on 2,020 IDSR instruction-response pairs. RAG-augmented Llama-2 achieved 97% accuracy with 75% reduction in memory requirements (8 GB VRAM).
NCST Malawi / Mzuzu University
Secured and led NCST-funded research on generative AI for equitable access to secondary education in Malawi. Designed full methodology, managed multi-person team, coordinated field data collection and model fine-tuning.
Mzuzu University
Built drought early warning model for Malawi using 40 years of NASA POWER and NOAA data. Applied ARIMA, STL, ETS, LSTM, XGBoost — selected XGBoost achieved R² 0.769, MAE 2.55. Integrated into interactive Power BI dashboard.
Mzuzu University
Co-supervising undergraduates building Chichewa NER corpus and model pipeline. Providing guidance on annotation schema design, AfroXLMR fine-tuning, inter-annotator agreement, and low-resource NLP methods.
Dept. of Applied Chemistry, Mzuzu University
Supported academic research through experimental data design, cleaning, and statistical analysis. Contributed to published study on ultrasound-assisted extraction from water hyacinth.
Mzuzu University, Malawi
Key Coursework: Research Methods, M&E, Statistics, Machine Learning, AI, Big Data Analytics, Predictive Modelling, Databases, Cloud Computing.
Mhango S.B., Kamlomo J.S., Sambito M.
Under review — preprint pending
Mhango S.B., Nyasulu C., Moyo R., Ndebvu S., Ngalande E., Mboma A.T., Peluffo-Ordonez D.H., Gonzalez-Vergara J.
Under review — preprint pending
Mboma A.T., Sibale B., Makombe D., Mwakilama E.P., Phiri H., Mhango S.B.
DOI: 10.21203/rs.3.rs-8988612/v1Ndebvu S., Phiri D., Mhango S.B., Phandera P., Nkhoma N.
DOI: 10.1002/isd2.70062Nababi J., Kamanula J.F., Alunga G.L., Wanjala F.P., Agnandji P., Nakawoombe M., Mhango S.B. et al.
DOI: 10.1016/j.biteb.2025.102370