██╗ ██╗ ██████╗
██║ ██╔╝██╔═══██╗
█████╔╝ ██║ ██║
██╔═██╗ ██║ ██║
██║ ██╗╚██████╔╝
╚═╝ ╚═╝ ╚═════╝
Building interpretable, resource-efficient AI for underserved communities — from Accra, Ghana 🇬🇭
I'm an AI/ML researcher and engineer driven by a core question: How can we build AI systems that augment human expertise in resource-constrained, high-stakes environments?
My work focuses on multimodal AI (vision + language + structured/temporal data), with particular emphasis on making systems that are interpretable, data-efficient, and actually deployable in the Global South.
Three research threads run through everything I do:
- 🔀 Multimodality & multilinguality — reasoning jointly across visual, linguistic, and temporal data; serving diverse linguistic communities
- 📉 Data scarcity & class imbalance — effective learning when labelled data is sparse or unevenly distributed, using LoRA adaptation, ensemble methods, and strategic augmentation
- 🔍 Explainability & trust — uncertainty-aware models that communicate their reasoning to domain experts making consequential decisions
| Challenge | Result | Core Approach |
|---|---|---|
| 🦠 SUA Outsmarting Outbreaks (Tanzania waterborne disease prediction) | 🥈 Silver Medal | Gradient Boosting + spatial indexing (cKDTree) fusing time-series health, climate, water quality & sanitation data |
| 🏥 Kenya Clinical Reasoning Challenge | 🥈 Silver Medal | LoRA-adapted Flan-T5 on 400 rural healthcare prompts; 5-fold ensemble; maternal health, paediatrics & critical care |
| 🌾 Ghana Crop Disease Detection Challenge | 🥉 Bronze Medal | YOLOv11 fine-tuned with stratified sampling; optimised for mobile inference in agricultural field conditions |
🏥 Healthcare AI → Clinical reasoning · Outbreak prediction · Medical imaging
💬 Multimodal & Multilingual NLP → Cross-modal reasoning · Low-resource languages · LLMs
📉 Learning under Scarcity → LoRA fine-tuning · Few-shot learning · Data augmentation
👁️ Computer Vision → Object detection (YOLO) · Edge-optimised CV · Anti-spoofing
🔍 Explainability & Trust → Uncertainty-aware systems · Interpretable ML · Causal inference
🌍 AI for Social Good → Responsible deployment · Accessibility · Developing-world AI
At KNUST, I served as a Trail Support Tutor and Academic Representative, organising interventions that supported over 2,000 students in core computing courses. I also led the Computer Engineering quiz team and ran hands-on machine learning tutorials for the KNUST Data Science and AI Club.
I'm always open to research collaborations on multimodal AI, healthcare AI, low-resource NLP, and any work building accessible technology for underserved communities.
📧 slyobeng111@gmail.com | 📍 Accra, Ghana | 🌐 i-ninte.vercel.app
"You cannot deploy a black box when lives are at stake."