Research
How can machine learning produce reliable, trustworthy decisions in environments where data is sparse, irregular and expensive to collect?
That question runs through everything below. It shows up as a forecasting problem in agriculture, as a monitoring problem in health, and as a coordination and safety problem in multi-agent systems.
Agricultural AI and remote sensing
My published work in this area began with a systematic review of deep learning for crop monitoring, screening 87 studies across 10 architecture families under PRISMA methodology, and mapping both reported performance and the methodological gaps behind it.
Current: do socio-economic variables improve yield prediction?
A controlled multi-modal experiment on Cameroonian agriculture. Three nested datasets — satellite only (D1), plus climate (D2), plus socio-economic indicators (D3) — evaluated on six deep learning architectures (CNN-LSTM, ConvLSTM, ConvLSTM-ViT, CNN-GRU, TCN, CNN-Transformer) across maize, cassava and cocoa in the Centre and Nord regions. The nesting is the point: it isolates the marginal contribution of each modality instead of reporting one black-box score.
Status: manuscript in preparation. Target venues: Computers and Electronics in Agriculture; Remote Sensing.
Supporting infrastructure
A Google Earth Engine pipeline extracting monthly NDVI, EVI, NDWI and NDRE indices together with Sentinel-1 radar indices, with a Landsat 5 → GIMMS fallback chain to close pre-2000 gaps. Alongside it, ensemble statistical methods under hard agronomic constraints to reconstruct and backfill incomplete regional production series.
AI for health
An IoT platform with embedded AI for patient follow-up in sub-Saharan Africa. The technical challenge mirrors the agricultural one: intermittent connectivity, irregular measurement intervals, and models that must stay useful when the input stream is incomplete rather than merely noisy.
Status: in development. [TO COMPLETE: partners, deployment context, current stage]
Multi-agent systems, robotics and AI safety
A longer-horizon strand of my work concerns collaborative AI: how multiple agents coordinate, what guarantees can be given about their joint behaviour, and how those guarantees hold up when the agents are embodied. AI safety is not a separate topic here — it is the question of whether a system that behaves well in evaluation continues to behave well in deployment, which is exactly the question that agricultural and health models raise in the field.
Collaborators
- Dayang Paul
- Moskolai Justin — University of Douala
- Ayissi Adolphe — University of Ngaoundéré
Open to
Doctoral supervision, research collaboration, joint grant applications, and reviewing in agricultural AI, remote sensing and applied machine learning. Get in touch.