Muhammed Noman

Built a Transformer-based text summarization model using TensorFlow with custom encoder-decoder architecture and attention mechanisms. Implemented positional encoding, multi-head attention, and sequence generation with decoder support. Trained on article-summary pairs and optimized for low-resource execution in Google Colab.

Built a deep learning model using Long Short-Term Memory (LSTM) networks to predict future stock prices based on historical time series data. Preprocessed financial data, engineered features, and trained the model using TensorFlow/Keras for sequence forecasting.

Developed a weather prediction model using Gated Recurrent Units (GRU) to forecast future temperature trends from historical weather data. Applied time series preprocessing, feature scaling, and built a GRU-based sequence model using TensorFlow/Keras.

Developed a poetry generator using TensorFlow/Keras, training a SimpleRNN model on the Gutenberg Poetry dataset to predict sequences. Engineered tokenization, embedding layers, and n-gram preprocessing to generate creative text outputs, demonstrating NLP and sequence modeling skills.

Implemented a Variational Autoencoder using TensorFlow/Keras to learn compressed latent representations of facial images. Trained the model on the LFW dataset to generate and reconstruct human faces with smooth latent space transitions.

Developed a deep learning model leveraging a CNN-based autoencoder to denoise 32x32 RGB images from the CIFAR-10 dataset. Implemented Gaussian noise injection, trained the model to reconstruct clean images, and visualized performance using TensorFlow/Keras, achieving robust noise reduction. Demonstrated proficiency in autoencoder architectures and image preprocessing for real-world applications.

Developed a Convolutional Neural Network (CNN) using TensorFlow/Keras for image-based garbage classification, capable of identifying six distinct waste categories (metal, trash, paper, glass, cardboard, plastic). Implemented data augmentation techniques and optimized the model for efficient and accurate waste sorting.

This project predicts whether a patient will be readmitted to a hospital within 30 days using classification algorithms like Logistic Regression, Decision Trees, Random Forest, and Gradient Boosting. The analysis includes data preprocessing, exploratory data analysis (EDA), and feature importance evaluation to identify key factors influencing readmission.

Developed a Convolutional Neural Network (CNN) using TensorFlow/Keras to accurately classify facial expressions into 7 categories. Implemented image preprocessing, data augmentation, and optimized the model architecture for robust performance in recognizing human emotions from images.

Developed a predictive model to forecast patient hospitalization duration using Python, Pandas, and Scikit-learn. Conducted exploratory data analysis (EDA), preprocessed data, and implemented regression algorithms (Linear Regression, Decision Tree, Random Forest) to achieve optimal performance. Evaluated models with metrics (MAE, MSE, R²) to identify key factors influencing LOS, aiding resource optimization in healthcare settings.

Grade: 3.5