CV
Education
- B.S. in Vehicle Engineering, Hefei University of Technology, 9/2012-6/2016
- M.E. in Mechanical Manufacture and Automation, Hefei University of Technology, 9/2016-6/2019
- Ph.D. in Mechanical Engineering, Missouri University of Science and Technology, 8/2019 - 5/2024
Work experience
- 2024 May.-Present.: Post Doctoral Associate
- University of Maryland, Department of Mechanical Engineering, College Park, MD, US
- Duties included:
- Designed a web-based application using TypeScript (frontend) and Python (backend) to enable online machine learning education.
- Collaborated with international companies to implement AI-driven solutions, improving process optimization.
- Developed proposals for projects with Siemens, Microsoft, and KU Leuven for NSF, NIST, and DOE.
- Developed similarity-based learning models using PHM datasets to predict equipment failures and optimize maintenance schedules.
- Implemented domain adaptation techniques with the CWRU dataset to enhance model generalization across diverse industrial settings.
- Taught Industrial AI course to PhD and Master students, covering AI models, PHM, and machine learning in smart manufacturing
- 2023 Jan.-Jul.: Machine Learning Engineer
- Peloton Interactive, New York, 441 9th Ave 6th, United States, 10001
- Duties included:
- Engaged in developing a cutting-edge system using AWS for Repetitive Action Counting (RepCount) and pose estimation in exercise routines, surpassing the state-of-the-art model by approximately 5%. This advancement provides users with in-depth insights into exercise patterns, enhancing progress tracking and workout optimization.
- Improved Repetitive Action Counting (RepCount) by 5% by training a 3D pose model on real-world action data from over 100 human subjects, using machine learning methods such as Swin Transformer and attention mechanism.
- Augmented the dataset by five-fold, employing techniques like flipping and rotation with OpenCV, enhancing data richness.
- Addressed challenges in RepCount like camera view changes, over-counting, under-counting, etc., and conducted data visualization.
- Boosted mAP by 8% in temporal action localization by designing a new point-level annotation with weakly-supervised learning.
- Technologies Utilized: Python, PyTorch, Tensorflow, Scikit-learn, AWS, Swin Transformer, attention mechanism, Google MediaPipe, OpenPose, etc.
- Supervisor: Solmaz Hajmohammadi (LinkedIn)
- 2019 Aug.- 2024 May: Graduate Research Assistant
- Missouri University of Science and Technology
Duties included: Engaged in pioneering research on human activity recognition, human-computer interaction (HCI), and human-robot collaboration (HRC) systems using sensor-driven big data, on-site customized cognition, and augmented reality (AR) control structures.:
- Recognition& Accuracy:
- Achieved 95% accuracy on instance segmentation of tools and parts using Mask RCNN, implemented object-oriented Python.
- Achieved 99% accuracy of offline multi-modal assembly action recognition using using CNN (VGG16, ResNet 50, and Inception V3) and RNN (LSTM and GRU) with CUDA/GPU programming.
- Attained 99% accuracy in real-time gesture recognition and 95% accuracy in NLP under diverse background noise condition using CNN and Google Speech Recognizer.
- Reached 95% accuracy on action recognition using inertial measurement unit (IMU) signals with TensorFlow and online assembly setp/sequence recognition using multi-view RGB camera inputs and fog computing.
- Implemented data augmentation through image processing and Generative Adversarial Networks (GANs) to amplify data from a limited dataset.
- Optimize motion patterns using unsupervised methods, such as the spectral cluster and Gaussian Mixture Model (GMM).
- System Development
- Crafted an eye gaze-based HCI system using Tobii tracker, OpenCV, PyTorch, PyQt, and Matplotlib.
- Designed an assembly action recognition and prediction system using Graph Neural Network (GNN) and YOLO.
- Constructed and analyzed skeleton/pose action dataset of human subjects using Google MediaPipe, SQL, Python, and Pandas.
- Designed an augmented reality (AR) assisted software interface for visual workforce training with CUDA/GPU programming.
- Initiated a proposal for YOLO-driven wound tissue detection software, utilizing GANs for data augmentation.
- Research & Mentorship
- Reviewed current machine learning literature to guide research directions and endeavors.
- Mentored eight undergraduate and master’s students in the research team.
- Produced and prepared over 10 presentations and research papers for publication with strong verbal and written communication skills.
– Technologies Utilized: Python, C++, TensorFlow, PyTorch, Keras, OpenCV, Numpy, CNN, RNN, GitHub, CUDA, ROS, etc.
- Recognition& Accuracy:
Skills
- Programming Languages: Python, C++, MATLAB, SQL, JaveScript, TypeScript
- Machine Learning Frameworks: TensorFlow, PyTorch, Keras, Scikit-learn
- Data Visualization & UI: Matplotlib, Pandas, PyQt
- AI & Computer Vision: Machine learning, Deep learning, Big data, Computer vision, HCI, HRC
- Model Architectures: CNN, RNN, GCN, VGG, ResNet, Inception, Mask-RCNN, YOLO, LSTM, GRU
- Cloud Computing: AWS EC2
- Development Tools: Anaconda, CUDA, GPU programming, Google Colab, PyCharm, VS Code, Jupyter, Spyder
- Operating Systems: Ubuntu, Linux, macOS, Windows
- Embedded Systems & Robotics: Raspberry Pi, Nvidia Jetson Nano, ROS
Journal Review Activities
• Engineering Applications of Artificial Intelligence (EAAI)
• IEEE Transactions on Circuits and Systems for Video Technology (TCSVT)
• Journal of Intelligent Manufacturing
• ASME Journal of Computing and Information Science in Engineering
• ASME Journal of Engineering and Science in Medical Diagnostics and Therapy
• ASME Journal of Mechanisms and Robotics
• ASME Open Journal of Engineering Journal of Intelligent Manufacturing Science China Technological Sciences
• ASME International Mechanical Engineering Congress & Exposition (IMECE) 2019 - 2023 Conference
• International Conference on Computer Science and Application Engineering (CSAE) 2019
Teaching
- Industrial Artificial Intelligence (ENME691, University of Maryland, Fall 2024)