Eliminate Pointless Packaging Using Computer Vision

Eliminate excessive packaging from online shopping with Social Media and Computer Vision.
I was the tech lead of the project in a group of 4 people.

• Communicated with the client effectively and translated his vision into a functioning product.
• Architectured a scalable web application infrastructure.
• Trained and deployed a Mask R-CNN model to Google Cloud Platform and leveraged several of Google’s services such as Cloud SQL, Cloud Storage, Compute Engine, and the Kubernetes engine.
• Implemented majority of the back end using the Django and Django REST Framework; front end using HTML, CSS, jQuery; and with features involving custom authentication system, infinite scrolling, and asynchronous AJAX file uploads.

Project Link: github.com/PointlessPackaging/pointlesspackaging

Parallel Seam Carving on GPU

Computer Vision Personal Project. CUDA and OpenCV.

• Implemented parallel version for parts of the Seam Carving algorithm. Exploited data and task level parallelism to compute energy images and maps.
• Utilized the OpenCV library for the CPU version of the algorithm and used CMake to manage the build process of the application.

Project Link: github.com/shafitek/Parallel-Seam-Carving-on-GPU


Machine Learning group project. Python, NumPy and Keras.

• Implemented a neural network based score function for the minimax algorithm with alpha-beta pruning based on an academic paper DeepChess.

• Generated 3+ million data points from a large dataset using parallel processing which reduced the generation time from hours to under 7 minutes.

Project Link: github.com/shafitek/DeepChess-AI