AI Engineer Intern at Viettel Post

Developed a Vietnamese NLP address-parsing system in Python using PhoBERT to map unstructured delivery addresses to standardized administrative divisions, achieving 87% parsing accuracy and reducing manual address correction. Deployed and optimized inference using NVIDIA Triton Inference Server, reducing inference latency by 27% for high-volume delivery requests. Built data preprocessing pipelines for 30,000+ noisy address records by handling abbreviations, spelling variations, and inconsistent formatting to improve robustness on real logistics data.

Undergraduate Researcher at Teletraffic Research Centre - The University of Adelaide

Researched and evaluated computer vision algorithms using Python and OpenCV for camera-based positioning with retroreflective tags across a dataset of 500+ images to support industrial ball mill fault monitoring. Applied projective geometry techniques to improve camera-based localization, increasing object tracking accuracy by 14% in experimental evaluation. Documented experimental findings and performance evaluations to support ongoing automated industrial fault-monitoring research.