- Sole contributor on a machine learning-based system for automatic license plate recognition developed by C++ , enabling law enforcement to track vehicles through a decentralized camera network.- Deploy and optimise machine learning models and license plate recognition on Windows/Linux servers using ShellScript, enabling data transfer and training from front-end cameras.- Sole developer of a website using D3.js to create an interface that allows users to tag objects in images by name. These tagged images are later used to train an AI model for object recognition.

- Top 500 in the globe, Quanta Computer is the largest notebook computer ODM company in the world with 10,000+ employees.- Contributed to the development of an automated testing framework leveraging Docker and Django to efficiently test notebook products.

- Trained a vehicle detection model using PyTorch for accurate object recognition. Optimized model performance and achieved faster inference by converting PyTorch models to TensorRT.- Reduced model size to accelerate inference speed, achieving over 10x faster computation while maintaining accuracy.- Independently developed a C++ system to process 30 real-time video streams at 15 FPS on an RTX 2080, deployable on both Windows and Linux.

- Independently collected and labeled data, and trained a pedestrian detection model using TensorFlow.- Optimized the model to enhance inference performance on ARM architecture, achieving over 5x faster inference speed.- Independently developed a facial recognition Check-in system and successfully implemented it in the company.