This repository showcases two machine learning projects developed by Hao-Yu Liao. Each project is implemented in a Jupyter Notebook and demonstrates practical applications of deep learning and generative AI.
File: detect_ai_vs_human_generated_images_Colab.ipynb
Platform: Built and trained on Google Colab
Competition Leaderboard: View on Kaggle — Search Howard Liao
Description:
This notebook was developed for the Detect AI vs. Human-Generated Images competition on Kaggle. It features a PyTorch-based ConvNeXT trained on Google Colab to classify whether an image is AI-generated or human-created.
Highlights:
- The base model is ConvNeXT, which I fine-tuned by freezing its core layers and adding custom classification layers for binary prediction (i.e., real vs. fake).
- Data preprocessing and augmentation
- Training and evaluation loop
- Achieved a top 7% solo ranking (41st out of 554 teams) in a Kaggle competition.
File: genai-personal-finance-advisor.ipynb
Platform: Built and tested on Kaggle Notebooks
Link: View on Kaggle
Description:
This notebook is part of a GenAI Capstone project. It implements a generative AI agent using the Gemini model to analyze financial data and provide decision-making support for buy, sell, or hold actions.
Highlights:
- Integration of LLM for financial reasoning
- Designed a financial system pipeline
- Generated actions based on historical price trends and up-to-date financial news
- Prompt engineering
Hao-Yu Liao
Email: liaohy.howard@gmail.com
GitHub: github.com/haoyuliao