Yousif Aldakoki
Selected Projects
IQ-CMAE — Multi-Modal Contrastive Masked Autoencoder
Self-supervised learning framework for wireless signal classification using raw I/Q traces. Combines masked autoencoding and contrastive learning with controlled multi-modal fusion.
Real-Time Waste Classification System
Cloud-assisted waste classifier using edge cameras and PyTorch inference. Achieved 96% accuracy with 87% lower inference cost.
MLP From Scratch (C++ / CUDA)
Implemented a neural network end-to-end without frameworks, including custom CUDA kernels.
Volatility Surface Modeling & Options Analytics
Modeled implied volatility surfaces from options data and analyzed dealer gamma/vega exposure.
Manufacturing Analytics Dashboard
Real-time PLC analytics platform with dashboards, editable tables, and statistical metrics.
Gyro-MNIST — Air-Drawn Digit Recognition
Embedded ML system where digits are drawn in the air using an Accelerometer and classified with Resnet-18 trained on MNIST-style representations.