Building and deploying deep learning systems — computer vision, sequence models, object detection and transformer-based NLP. Portfolio includes implementation, training, evaluation, and deployment-ready artifacts.
Powerful technologies used to build intelligent deep learning solutions
Explore implementations across neural network architectures and deep learning models
Dataset: MNIST Dataset
Basic feed-forward ANN classifier for handwritten digit recognition with dropout and batch normalization.
Dataset: Telco Churn
ANN for predicting customer churn using structured datasets, feature engineering, and precision-recall evaluation.
Dataset: CIFAR-10
Custom CNN architecture for image classification with data augmentation, LR scheduling, and feature visualization.
Dataset: ImageNet
Fine-tuning pretrained models including ResNet50, MobileNetV2, and EfficientNet for multi-class classification.
Dataset: Custom Dataset
YOLOv8 detection pipeline with training, evaluation, dataset prep, and real-time inference integration.
Dataset: Time Series Data
Multi-layer LSTM for future price prediction using sliding window preprocessing and backtesting evaluation.
Dataset: English–French
Encoder–decoder LSTM with Bahdanau attention, teacher forcing, and beam search decoding.
Dataset: News Articles
Summarization using T5/BART transformers with fine-tuning, ROUGE evaluation, and custom training loop.