Case studies say what I did, what the team did, and what the evidence supports. Smaller experiments and learning work are kept separate, so they're easy to weigh for what they are.
Smart hydroponics connecting sensors, embedded devices, cloud services, and a web interface.
My contribution. I founded the initiative and designed the end-to-end software and data architecture — backend services, REST APIs, sensor-data pipelines, monitoring workflows, and the interfaces on top of them.
Drone imagery and deep learning for detecting transmission-tower components.
My contribution. As one contributor on a collaborative project with SEM Ltd., I worked on data preparation, model training and evaluation, edge deployment to an NVIDIA Jetson Orin Nano, DJI drone integration, and system-level validation.
Comparing four architectures for pneumonia detection from chest X-rays.
My contribution. I designed and evaluated the models, compared DenseNet121, ResNet50, EfficientNet-B3, and a custom CNN, addressed class imbalance and overfitting, tuned the decision threshold, and deployed the selected model behind a Streamlit interface.
PX4 SITL, ROS 2, and Gazebo for autonomous control and telemetry validation.
My contribution. I developed and validated an autonomous drone-control simulation, implemented real-time telemetry and OFFBOARD control workflows, validated IMU, GPS, and positioning data, and performed system-level debugging and log analysis.
A theme-native build-your-own-bundle experience for Shopify Online Store 2.0.
My contribution. I built a custom bundle product template and section on top of Shopify's Dawn 11 theme, driven by merchant-editable metafields, with validation that prevents incomplete or duplicate bundles.
A multiplayer stock-trading game with live NYSE prices, portfolios, and admin-run games.
My contribution. On a team project for a graduate software course, I delivered GUI improvements, a portfolio performance visualization, a reCAPTCHA login fix, and a new admin game-management page during Sprint 3.
An explainable image-processing pipeline for extracting and measuring leaf vein networks.
My contribution. Co-authored with Annur Anika for ENGI 9804 (Image Processing and Applications) at Memorial University. We built a transparent pipeline in which every processing decision can be inspected and measured.
Python
OpenCV-style image processing
PCA
K-Means
Random Forest
Read the case study
Experiments and learning work
Smaller prototypes, customizations, and coursework. Described from their repositories; no outcomes beyond what those record.
Academic project
CNN–LSTM Image Captioning
VGG16 image features feeding an LSTM language model that predicts captions word by word. A prototype encoder–decoder combining VGG16 feature extraction with an LSTM, demonstrating transfer learning, tokenization, and sequence modelling.
A Python and Flask simulation of soil-moisture sensing, pump control, and a small dashboard. A software-only prototype of the control logic and monitoring interface, built as a safe step before Raspberry Pi GPIO hardware.
A source-controlled Dawn 11 workspace for storefront customization and experiments. Keeps theme work reviewable in Git and provides a base for testing Liquid, theme settings, and responsive storefront behaviour away from production.
A YouTube-inspired layout recreated with Vue 3, Quasar 2, and Vue Router. Recreates a toolbar, navigation drawer, search layout, and video cards with dynamic view counts — frontend only, with no production backend.
Completed notebooks from Python fundamentals through NumPy, Pandas, and a Netflix dataset analysis. Two repositories from CKPCET workshops, including an exploratory analysis of a 6,234-title Netflix dataset.