Ameyaw Emmanuel Kakari

I'm developing QuantLab, a hardware-aware mixed-precision quantization framework for transformer models, evaluating accuracy, latency, and memory trade-offs through GPU benchmarking.

My independent research investigates efficient deep learning through model compression, quantization, computational efficiency, and deployment on modern hardware. I build reproducible GPU-accelerated experiments to study how numerical precision affects accuracy, memory footprint, inference performance, and hardware behavior.

I also built an explainable multimodal medical AI system for skin cancer risk stratification, combining EfficientNet-B0 image embeddings with structured patient metadata. The work includes Grad-CAM++, SHAP, Monte Carlo Dropout, temperature scaling, fairness auditing, and patient-level stratified cross-validation.

As a Computer Science undergraduate, I'm focused on machine learning engineering, efficient AI, computer vision, explainable AI, natural language processing, and machine learning systems. I also lead a student robotics community that connects programming, embedded systems, engineering, and hands-on technical learning.