Soyeb Hoque

I’m a B.Tech in Information Technology student focused on building production-ready ML systems, especially in computer vision and applied deep learning. I work end-to-end—from preprocessing and model design to evaluation and deployment-ready pipelines.

My projects include an Advanced Maize Disease Classification system where I achieved 99.44% accuracy on 5-class classification using MobileNetV2 with CBAM attention, plus a hybrid CNN-SVM approach that outperformed softmax by 2.28%. I also built real-time face detection with OpenCV and Haar cascades (95%+ accuracy across varied lighting) and an end-to-end OCR pipeline with 92%+ accuracy using text region detection and character recognition.

I’ve extended my work to satellite image analysis using ensemble methods and transfer learning for terrain/land cover tasks, and I’ve implemented 15+ supervised learning algorithms from scratch with k-fold cross-validation and GridSearch tuning. I’m actively excited by impact-driven research and competition work—I’m an ISRO hackathon finalist, a top 1% Kaggle contributor, and I maintain 55+ GitHub repositories.