McGill FAIM Hackathon - Portfolio Optimization

A comprehensive quantitative trading strategy developed for the McGill Finance and Investment Management (FAIM) hackathon.

Team & Leadership

Led a team of 5 in developing and implementing the trading strategy, coordinating across:

• Data engineering
• Model development
• Portfolio management
• Performance analysis

Technical Approach

Stock Prediction Models

• Elastic Net Regression: For feature selection and prediction
• LSTM Neural Networks: For capturing temporal patterns in price movements
• Trained on 147 investment factors alongside seasonal data

High-Performance Computing

Leveraged Queen's University compute cluster for:

• Multi-GPU training optimization
• Large-scale backtesting
• Parallel hyperparameter search

Portfolio Management

Implemented Black-Litterman portfolio management:

• Managed 50-100 stocks in long and short positions
• Combined quantitative signals with market equilibrium
• Dynamic rebalancing based on model predictions

Results

Achieved an alpha of 0.00137, demonstrating positive risk-adjusted returns above market benchmark.