Portfolio
Featured AI Projects
Cutting-edge machine learning and deep learning solutions across healthcare, computer vision, NLP, and IoT domains
Similarity Search Engine
Rapid retrieval system using pretrained ResNet for feature extraction, PCA dimensionality reduction (2048→150 features), and KNN for lightning-fast nearest neighbor search.
Chest X-Ray Pneumonia Detector
Developed a pneumonia detection system using MobileNet-V3 with transfer learning, achieving 94% accuracy on chest X-rays. Optimized via knowledge distillation for edge devices and deployed on AWS SageMaker.
Signature Verification System
Few-shot learning solution using Siamese networks for robust signature verification with minimal training samples. Ideal for secure authentication and fraud prevention systems.
OCT Retinal Disease Classification
Developed advanced MLP mixture model achieving 98.4% accuracy for multi-class retinal disease classification using OCT images, with weighted loss functions to address severe class imbalance.
Liver Segmentation with U-Net
Optimized medical image segmentation using U-Net and Half U-Net architectures, achieving 97.6% IoU and 94% Dice Coefficient with custom loss functions for precision healthcare imaging.
IoT Rash Driving Detection
Secure IoT system using ESP32 for real-time rash driving detection and alerts. Integrated multiple sensors for comprehensive vehicle behavior monitoring and safety enhancement.
Biomedical Entity Extraction
Built production-ready NER system using BioBERT for extracting biomedical entities (diseases, treatments, genes) from clinical texts, providing structured healthcare data insights.
Quora Question Pair Similarity
Developed a question similarity detection system to identify semantically similar Quora questions and tag shared answers using EDA, TF-IDF, Word2Vec, and ensemble models.