Anh-Tu NGUYEN

* Develop and industrialize computer vision models (detection, segmentation, tracking) for an ADAS software deployed on Ford trucks, achieving a 90% reduction in sensor costs. * Implement and deploy adaptive depth estimation models for obstacle avoidance in robots and drones; co-inventor of an international patent. * Implement detection and tracking models for VRUs (Vulnerable Road Users). * Implement and optimize camera calibration algorithms, achieving < 0.1 pixel error. * Industrialize algorithms: pruning and quantization (-80% execution time), sim-to-real adaptation (20% improvement in robustness and stability). * Develop internal tools for automated data processing and algorithm validation. * Set up an MLOps system. * Implement processing and fusion pipelines for 3D vision data (LiDAR, ToF, RGB-D). * Deploy neural networks on embedded platforms: Qualcomm, NVIDIA, Luxonis, Ambarella.

* Develop and productionize models for damage detection and segmentation, OCR, and car parts segmentation. * Design a damage size estimation algorithm achieving <5% error. * Set up an MLOps stack (DVC, MLflow, Kubernetes, Prometheus, Grafana). * Design and implement a REST API for ML model inference, deployed on AWS. * Set up a centralized system for logging, metrics, tracing, and alerting.

* Designed and implemented a scalable soil carbon database, modeling geospatial and environmental data using PostgreSQL/PostGIS. * Architected and built robust ETL pipelines and a datamart to aggregate, transform, and serve data from heterogeneous sources. * Developed a web-based data access layer (Symfony) exposing structured data to end users via performant queries and APIs. * Contributed to system design decisions around data modeling, storage optimization, and query performance. * Tech stack: PHP (Symfony 3), JavaScript, PostgreSQL (PostGIS), Oracle SQL Developer.