- Hello, I’m Jason Fu.
- I’m interested in reading books and learning new technologies.
- I’m currently working as a Machine Learning Senior Manager in Standard Chartered Bank.
- My expertise is Machine Learning, Deep Learning, ML Ops, Computer Vision etc..
This repository primarily comprises real-world use cases and applications of AI/ML in the banking industry. It includes both projects I have personally undertaken and those inspired by collaborative efforts within the industry.
Throughout my experience in various roles within the banking industry, such as personalized marketing strategy, risk management, fraud detection, and robotic process automation (OCR), I have witnessed the transition from traditional statistical modeling to AI/ML modeling. Banks have been increasingly adopting advanced AI/ML techniques to keep pace with the acceleration of cloud migration and the rapid accumulation of data. Utilizing these advanced technologies will be crucial in setting themselves apart from competitors.
The repository covers AI/ML applications in the following areas:
- Automated feature engineering
- Semi-auto feature selection techniques
- Making sense of time series features
- Missing value imputation using DataWig
- Tactics for imbalanced data
- Extract model insight using permutation importance
- Checkbox detection on bank document
- Signature detection on scanned documents using MMdetection
- Bank form structure and contents recognition
- Using OCR to extract contents from bank document
- Intelligent Document Understanding Visual Questions Answering
- Anomaly detection for credit card fraud
- Unsupervised anomaly detection for card fraud
- Credit card default prediction using LGBM method
- Company bankruptcy prediction
- Marketing Campaign Effective Targeting
- End to End AutoML for insurance cross selling
- Application of Customer Segmentation
- Use FastAPI to deploy ML model into production
- Flask API web services for ML model deployment
- ML System Production in Google Cloud
- ML application deployment using OpenShift container
- Log, Exception and Code Standards for production deployment
- ML Ops CICD
- File structures for deployment ready scripts
The implementation of AI/ML in the banking industry encounters numerous obstacles, such as challenges related to integrating with legacy systems, ensuring compliance with strict regulations, investing in infrastructure setup, addressing data governance and security concerns, and more. This repository is an ongoing project, and I welcome any feedback or comments to further enhance its content. Feel free to contact me at sfu012@sina.com.