Past projects.
Built along the way.

Built packet labelling, bidirectional flow construction and feature extraction in Python. The case study explains the flow-based detection design and the source excerpts preserved in the repository.

The public code covers preprocessing; it does not include a reproducible end-to-end classifier or measured detection results.

Encoded an 8 × 8 board with four state channels, defined rewards and built a Keras network to estimate four directional action values.

The public archive preserves the network and test routine. The full training loop and saved model are not included, so path-finding performance cannot be reproduced from the repository alone.

Course project · Zero To Mastery (ZTM)

Explored the UCI Cleveland heart-disease dataset in a Python notebook, comparing Logistic Regression, K-Nearest Neighbours and Random Forest against a binary target.

The work covers exploratory plots, a train/test split, randomised and grid searches, and five-fold cross-validation. I used ROC curves, confusion matrices, precision, recall and F1 to examine predictions, then inspected the logistic-regression coefficients.

Course project · Zero To Mastery (ZTM)

Built a regression notebook around Kaggle’s Bluebook for Bulldozers dataset to estimate equipment sale prices from machine attributes and previous auction records.

Extracted calendar features from sale dates, handled missing values and encoded categories before training a Random Forest regressor. Used 2012 sales for validation, compared RMSLE, MAE and R², explored feature importance and exported test-set predictions.

Course project · Zero To Mastery (ZTM)

Built a multi-class image-classification notebook for 120 dog breeds using Kaggle’s Dog Breed Identification dataset and a pretrained MobileNetV2 model from TensorFlow Hub.

The pipeline resizes images to 224 × 224 pixels and batches them with tf.data. A softmax classification layer produces breed probabilities; TensorBoard and early stopping support training, with prediction visualisations and CSV export completing the workflow.

Created a menu-driven address book as a degree course project, using classes and file storage to organise contact and business records.

The project supports creating, searching, displaying, modifying and deleting entries. I extended the search workflow with a “most searched” feature. The README illustrates hospital records and the main menu, while the repository preserves a source-code excerpt and sample data files.

Developed a Java desktop prototype for exploring restaurant orders and customer waiting times, with a Swing interface and first-come, first-served queue logic.

The source tracks item quantities, customer arrival times, service times and waiting times. Preparation-time calculations feed the queue model, making the connection between an order and the delay experienced by the next customer visible in the program.

Built a shopping-site project with category pages, product details and a cart interface, using a Bootstrap-based storefront with PHP and MySQL workflows.

The repository includes registration and login, adding products to a cart, order confirmation and order history. PHP sessions connect customers to those workflows. The hosted preview shows the static storefront; the PHP application is available in the repository.

Created an Android demonstration app to explore Ionic interfaces and native device features through Cordova. The repository includes screenshots, the packaged app and bundled source maps.

The code covers profile forms, camera and photo-library access, geolocation, network-status events and social sharing. A local SQLite example adds and retrieves developer records, bringing interface components and device integrations into one project.