- Architected and led the development of DocAI, an AI-powered document management system with free-template OCR and document understanding capabilities.- Led AI/ML development and architecture for TADA, a low-code data platform accelerating complex data pipeline construction. The main contributor of the platform's SDK which provides users blocks of codes (called Components) that are tangible.- Provided expert advisory services for diverse customer projects.- Developed Mapping and… Show more - Architected and led the development of DocAI, an AI-powered document management system with free-template OCR and document understanding capabilities.- Led AI/ML development and architecture for TADA, a low-code data platform accelerating complex data pipeline construction. The main contributor of the platform's SDK which provides users blocks of codes (called Components) that are tangible.- Provided expert advisory services for diverse customer projects.- Developed Mapping and Localization module for the auto-parking system on trucks, achieving centimeter-level precision. Show less

I started to work in VinAI following the program "Talent Vingroup AI Engineer Nurturing" of Vingroup. Working here has given me a great opportunity to join the Autonomous Vehicle project and work with great people.And below are the achievements that I got during the time here:- Understood, improved, and deployed successfully Autoware.Auto on a Lincoln MKZ.- Working with Lidar which consists of investigating the technology of kinds of Lidar, researching 3D Object Detection using… Show more I started to work in VinAI following the program "Talent Vingroup AI Engineer Nurturing" of Vingroup. Working here has given me a great opportunity to join the Autonomous Vehicle project and work with great people.And below are the achievements that I got during the time here:- Understood, improved, and deployed successfully Autoware.Auto on a Lincoln MKZ.- Working with Lidar which consists of investigating the technology of kinds of Lidar, researching 3D Object Detection using Lidar, converting the driver for a Lidar from ROS1 to ROS2, and developing synchronization between Lidar and Camera.- Worked in a full flow of generating HD Map to enable the car's navigation system from collecting data to make the map work on the vehicle. The map achieved centimeter precision. The article writing about our autonomous vehicle using my team's map is attached below. Show less

VInBigData is the place where hosted the program "Talent Vingroup AI Engineer Nurturing" and I joined this program to update and review my knowledge, as well as have a chance to discover the Northern of Viet Nam.Here, I focused on Action Recognition by using Deep Learning. Below is what I achieved:- Feature engineering the skeleton to feed in XGBoost for action recognition and got the result of 81% accuracy.- Have a deep understanding and successfully trained VPN (a SOTA action… Show more VInBigData is the place where hosted the program "Talent Vingroup AI Engineer Nurturing" and I joined this program to update and review my knowledge, as well as have a chance to discover the Northern of Viet Nam.Here, I focused on Action Recognition by using Deep Learning. Below is what I achieved:- Feature engineering the skeleton to feed in XGBoost for action recognition and got the result of 81% accuracy.- Have a deep understanding and successfully trained VPN (a SOTA action recognition model) with an accuracy of 96%. Show less

I worked in a team of 2 people and the mission of the AI team in Kobiton is to develop AI models to serve the ITA (Intelligent Testing Automation) of Kobiton. In detail, ITA helps users simulate a sequence of actions of a mobile tester on multiple devices' models. During my time here, I achieved the following:- Benchmark and speed up the performance of AI service by 10 times.- Worked in a full flow of Data Science project including defining data annotation methods, developing… Show more I worked in a team of 2 people and the mission of the AI team in Kobiton is to develop AI models to serve the ITA (Intelligent Testing Automation) of Kobiton. In detail, ITA helps users simulate a sequence of actions of a mobile tester on multiple devices' models. During my time here, I achieved the following:- Benchmark and speed up the performance of AI service by 10 times.- Worked in a full flow of Data Science project including defining data annotation methods, developing data crawlers, developing annotation interfaces, data processing, training, fine-tuning models, deploying models, and developing back-end server to provide AI APIs.- Model reached an accuracy of 97% for matching task.- Worked remotely with CTO in the USA to research and develop features. Show less

I led a 5-members team to develop a surveillance system for smart factories. We developed and improved the human-counting features with YOLO v3 by conducting experiments with Faceboxes and integrate Tracker models such as SiamMask and THOR. Besides that, one of my colleagues and I also developed a website to stream the cameras in a factory.

I developed a face matching system by FaceNet and DBScan independently. After that, my colleagues and I did research on MTCNN and developed a forbidden-arena warning features for the surveillance system of Vietnam Unilever Factory.

I worked on CNNs architecture to solve classification and object detection problem. In detail, we applied Faster-RCNN by Tensorflow Object Detection API and Xception by Keras into a system to detect objects on magazines.