Role and Responsibilities:- Deep dive in design architecture for video classification, Yolo detection (and variants) integration on edge devices with a multitask lack of datasets.- Directly working in the Research and Development of Generative AI applications in the Furniture Design domain.- Engage in discussions with stakeholders departments to identify and resolve bottlenecks in project workflows. - Sharing insights and integrating feedback from various contributors and customers to enhance project outcomes.

Roles and Responsibilities- Take responsibility for the ML lifecycle project in NLP domains.- Collaborate with managers to Research and Develop foundational models for multitasking in AI applications.- Demonstrates high adaptability to continuously evolving technologies in AI research, leveraging SOTA techniques to enhance product performance.- Led a team of 6 engineers in NLP fields (Conversation Chatbot, Task-Oriented Domain, Asymmetric search, and Translation) from research to product.Key Accomplishments- Contributed to the development of the Mina chatbot.- Build an Asymmetric system: Retrieve and Rerank modules (data, architecture, and training) with bilingual (English and Vietnamese).- Harnessing the potential of graph neural networks and graph of thought on reasoning.Technologies- Developments: Pytorch-lightning, Tensorfow, Jax, Huggingface.- Track & Monitoring: Wandb, MLflow, Neptune, etc.- Optimizer: ONNX, Openvio and TensorRT.

Smart Fashion AI- Design a FashionGAN model: Generator (Unet + Inceptionv4) transfers the patterns (details cloth, image, text) and Discriminator (VGG19 + Neural Style Transfer) keeps the style color features. - Fully supported research field: Segmentations, Human pose, pattern reconstruction- Collected and processed data using Image Processing techniques to handle missed inputs and outputs. - Evaluating the prediction and controlling code's version using GIT

Kinship Face in the Wild:- Kinship verification aims to determine whether there is a kin relation between a pair of given face images.- There are four representative types of kin relations: Father-Son (F-S), Father-Daughter (F-D), Mother-Son (M-S) and Mother-Daughter (M-D), respectively.- In this study, we apply the Local Binary Pattern (LBP) family and Histogram of Gradient (HOG) descriptors used for image extracting features.- Propose the multi-color variant spaces combined with the features selection (Fisher Score, Constraint Score) to find relevant and irrelevant features after that ranking and normalize embeding. - Finally, the performance will be evaluated using the Support Vector Machine (SVM). - My study has been published in two papers at two international conferences