Nguyen Phuc

Experience

• 

  ESilicon

  Mar 2015 - Apr 2017

  IP Design Engineer
• 

  FAU Erlangen-Nürnberg

  Nov 2020 - Sept 2021

  Using the info set from: https://u.cs.biu.ac.il/~koppel/BlogCorpus.htm I build an enquiry and ranking system for taking a question as input and retrieving all the blogs that contain the terms and also are ranked.The query are provided as an English query joined does on google search.The query could contain sentence/keyword/phrases query.The queries may not contain the precise words as presented within the blogs. Relevant words and similar word matches are taken into consideration.There are several steps in this project.Github link: https://github.com/ntphuc0101/Semantic-Search.git1. Data Cleaning and Pre-processing.2. Using BERT to embedded paragraphs of papers using bert-base-nli-mean-tokens pretrained model.3. Find the closest 5 sentences of the corpus for every query sentence supported cosine similarity. Show less My primary focus revolves around crafting a classification algorithm tailored to the software engineering domain. This algorithm is designed to deduce the type of code or programming language (anatomy) and the intended purpose (viewing direction) from a given software source code snippet.Throughout this project, I've undertaken the successful implementation of diverse machine learning model architectures. These include adapting established models such as Densenet, Pre-Trained Densenet Models trained on extensive code repositories, and leveraging Pre-Trained VGG16 models fine-tuned on software-specific datasets.The outcomes of these model implementations have been exceptionally positive, consistently achieving classification accuracy rates exceeding 90%. This research has significant implications for improving code analysis and understanding in the software engineering field. Show less There are two ways to improve the performance of Convolutional Neural Network. We can either use Aging evolution or scaling a convolutional neural network. Aging evolution, a variant of tournament selection by which genotypes die according to their age, favoring the young. This can help us create a good base-line network with good performance. To further improve the accuracy, then Proposed Compound Scaling for Convolution Network Scaling is applied. Both these algorithms can help reach the higher performance of models.Here are my report and video. Show less

  • Semantic-Search

    Apr 2021 - Sept 2021

  • Intern

    Nov 2020 - Apr 2021

  • Voice Activity Detection

    Mar 2021 - Mar 2021

  • Seminar Scaling Convolutional Neural Network

    Nov 2020 - Nov 2020
• 

  Fraunhofer Institute for Integrated Circuits IIS, Division Engineering of Adaptive Systems EAS

  May 2021 - Feb 2022

  Student Assistant

  In this role, I led the design and implementation of cutting-edge image and video compression algorithms, resulting in a 30% performance improvement. I conducted system tests, enhanced risk coverage, and provided support to colleagues while delivering research presentations.Skills: python, C++, Unittest, CI/CD github, CMake, VCPKG, Conan , testing with Catch2

• 

  Fraunhofer IIS

  May 2021 - Feb 2022

  Masterarbeit
• 

  DC Vision Systems GmbH

  Feb 2022 - now

  Software Engineer 3D reconstruction

  Develop software for computer vision, embedded systems, Image processing systems, algorithms, SLAM. C++, Python, UNIX, Clang-Tidy, Conan