Dũng Anh

Built regression and classification models to predict technician ETA and job duration, achieving a 23% improvement in prediction accuracy through strategic model selection and feature engineering. Collaborated on model export to ONNX and provided Python inference wrappers for seamless integration by backend teams.

Developed multilingual speech recognition and vehicle anomaly detection models for an automotive IoT platform, achieving 92% accuracy on multilingual command recognition and 94% precision in anomaly detection. Managed end-to-end ML workflow including data ingestion, feature extraction, model training, evaluation, and delivery of optimized inference artifacts.

Developed predictive maintenance models for industrial CNC machinery, achieving 91% accuracy in failure prediction with 72-hour advance warning capability. Reduced false alarm rates by 35% through systematic model selection and ensemble techniques, contributing to an 18% reduction in unplanned downtime costs.

Developed demand forecasting and recommendation systems for an e-commerce platform, achieving a 25% improvement in demand forecast accuracy and a 32% increase in recommendation click-through rates. Owned end-to-end ML processes including data wrangling, feature creation, model training, validation, and handoff for integration.

Developed predictive health monitoring models for continuous patient vital signs analysis from IoT medical devices, achieving 94% sensitivity in early disease detection. Led core ML workflow including data consolidation, feature crafting, model prototyping, rigorous evaluation, and handoff of inference artifacts.

Studied Information Technology at FPT University. Principal author of the He LiGCN - GCN model presented at ICAA 2025.