Machine Learning & AI Engineer
Building practical machine learning systems and applications that connect model development, evaluation, and usable software.
I am a Software Engineering graduate from Zhengzhou University focused on applied machine learning, deep learning, and computer vision. My work spans data preparation, model training and evaluation, explainable AI, and integrating model-driven functionality into applications and APIs. I also build full-stack software to support practical workflows around these systems.
An end-to-end equipment-risk analysis platform built around a machine learning pipeline, with model explanations and an interactive dashboard.
- XGBoost-based failure-risk prediction with SHAP explanations.
- Jupyter workflows for data auditing, preprocessing, model comparison, evaluation, and explainability.
- FastAPI inference services and a Next.js dashboard for individual predictions, batch CSV analysis, fleet views, and simulated telemetry.
Stack:
A CPU-focused image-classification project that distinguishes NORMAL and PNEUMONIA chest X-rays and provides visual explanations of model attention.
- ResNet18-based classification using PyTorch and Torchvision.
- Grad-CAM visualizations and documented evaluation on a held-out test set.
- FastAPI backend and Next.js dashboard, supported by separate training and evaluation notebooks.
Stack:
Educational and research project only; results are not evidence of clinical effectiveness and must not be used for patient care.
A CNN-based image-classification application for identifying plant leaf conditions across 38 disease and healthy-leaf classes covering 14 plant species.
- Image preprocessing and model training with TensorFlow/Keras.
- Training visualizations, a confusion matrix, and classification-report workflows.
- Streamlit interface for image upload and prediction.
Stack:
A real-time webcam project that detects and classifies a limited set of hand signs representing the ASL letters I, L, O, V, E, and U.
- Hand-region detection and preprocessing with OpenCV and MediaPipe-based tooling.
- TensorFlow/Keras classification using collected image samples.
- Live webcam inference with predicted labels displayed on screen.
Stack:
A webcam-based attendance application that identifies registered faces and records recognized individuals with timestamps.
- Face detection and matching against registered face images.
- Attendance records stored in CSV format.
- Visual recognition feedback and handling for unknown faces.
Stack:
These projects demonstrate complementary experience in application architecture, backend development, database-backed workflows, and user interface engineering.
| Project | Description | Repository |
|---|---|---|
| PharmacyPOS (PharmaCore ERP) | Pharmacy operations platform covering point of sale, inventory, batch and expiry tracking, role-based access, branch workflows, and reporting. | View project |
| Restaurant POS (EzPOS) | Restaurant management and point-of-sale application covering orders, table service, menu management, stock tracking, payments, and reporting. | View project |
| Customer Feedback Dashboard | Feedback collection and rating dashboard with PostgreSQL persistence, Chart.js visualizations, and keyword-based sentiment classification. | View project |
Focused on machine learning, computer vision, and practical AI engineering.
