使用说明
Machine Learning Portfolio
A collection of machine learning, deep learning, computer vision, and generative AI projects developed through academic and hands-on project work.
This portfolio demonstrates practical experience across time-series forecasting, neural networks, computer vision, natural-language interfaces, SQL generation, data preprocessing, model evaluation, and interactive machine-learning applications.
Projects
1. Smart Waste Management – Time-Series Forecasting
A deep-learning project focused on forecasting waste-bin fill levels from time-series sensor data for smart waste-management applications.
The project explores multiple neural-network architectures and compares predicted bin levels against actual sensor observations.
Key Areas
- Time-series preprocessing
- Sensor-data analysis
- Feature and sequence preparation
- LSTM neural networks
- CNN-based sequence modeling
- Hybrid CNN-LSTM architectures
- Model training and evaluation
- Forecasting and prediction analysis
Technologies:
Python, TensorFlow/Keras, Pandas, NumPy, Scikit-learn, Matplotlib
2. Farm Insect Classification Using YOLOv8
A computer-vision application for identifying agricultural insects from uploaded images using a YOLOv8 classification model.
The project includes model training and an interactive Streamlit application where users can upload an insect image and receive the predicted insect class with confidence scores.
Key Areas
- Computer vision
- Image classification
- YOLOv8
- Transfer learning
- Image preprocessing
- Model inference
- Confidence-based predictions
- Interactive ML application development
Technologies:
Python, YOLOv8, Ultralytics, PyTorch, OpenCV, Streamlit
3. BridgesAI – NLP-to-SQL Bridge Infrastructure Explorer
BridgesAI is a natural-language analytics application that allows users to ask questions about bridge infrastructure data without manually writing SQL.
The application interprets a natural-language question, determines its analytical intent, retrieves relevant database schema information, generates a safe SQLite query using an LLM, executes the query, and presents the results through an interactive Streamlit interface.
Core Pipeline
Natural-Language Question
|
v
Domain Relevance Validation
|
v
Intent Classification
|
v
Geographic Entity Detection
|
v
Schema-Aware RAG
|
v
LLM SQL Generation
|
v
SQL Safety Validation
|
v
SQLite Execution
|
v
Results & Visualization
Key Areas
- Natural Language Processing
- NLP-to-SQL
- Retrieval-Augmented Generation (RAG)
- OpenAI embeddings
- LLM integration
- Intent classification
- Schema-aware retrieval
- SQL generation and validation
- Geographic entity matching
- Interactive data visualization
Technologies:
Python, Streamlit, OpenAI API, SQLite, SQLAlchemy, Pandas, NumPy, RapidFuzz
Technical Skills Demonstrated
Machine Learning & Deep Learning
- Time-series forecasting
- LSTM neural networks
- CNN and hybrid CNN-LSTM architectures
- Computer vision
- YOLOv8
- Transfer learning
- Model training and evaluation
Generative AI & NLP
- Large Language Model API integration
- Retrieval-Augmented Generation (RAG)
- Embeddings and semantic similarity
- Natural-language-to-SQL generation
- Intent classification
- Prompt engineering
Data & Analytics
- Python
- Pandas
- NumPy
- SQL
- SQLite
- SQLAlchemy
- Data preprocessing
- Data validation
- Exploratory analysis
- Data visualization
Application Development
- Streamlit
- Interactive ML applications
- Model inference workflows
- Environment-based API key management
- Git and GitHub
Project Overview
| Project | ML Area | Primary Application |
|---|---|---|
| Smart Waste Management | Deep Learning / Time Series | Waste-bin fill-level forecasting |
| Farm Insect Classification | Computer Vision | Agricultural insect classification |
| BridgesAI | Generative AI / NLP | Natural-language data analytics |
Repository Structure
ML_Portfolio/
│
├── README.md
│
├── smart-waste-management/
│ ├── README.md
│ ├── src/
│ ├── data/
│ └── results/
│
├── farm-insect-yolov8/
│ ├── README.md
│ └── ...
│
└── bridge-nlp-to-sql/
├── app.py
├── README.md
├── requirements.txt
└── data/
Additional Projects
Additional machine-learning, data science, and AI projects will be added as they are prepared for public portfolio presentation.
About This Portfolio
This repository serves as a consolidated portfolio of my machine-learning and AI work.
The projects include academic and hands-on implementations covering different stages of the machine-learning lifecycle, including data preparation, model development, training, evaluation, inference, natural-language interaction, and interactive application development.
The portfolio is intended to demonstrate both foundational machine-learning knowledge and practical experience building end-to-end data and AI applications.
