Core Technical Capabilities

​

1. Portfolio Construction and Tracking

• Constructs long and short equity portfolios across a defined range of risk levels.

• Supports multiple markets and currencies; currently covers approximately 9,000 equities across NASDAQ, NYSE, and AMEX (as of July 2025).

• Tracks portfolio performance with ongoing updates based on market data.

​

2. Multi-Layer Data Integration

• Market Data: End-of-day prices, trading volumes, dividends, and stock splits.

• Fundamental Data: Quarterly financial statements, company attributes, and sector classifications.

• Historical Coverage: 16 years of historical data, used for backtesting and covariance estimation.

​

3. Risk and Performance Metrics

• Computes more than 30 standard quantitative metrics, including Beta, Sharpe Ratio, Sortino Ratio, covariance matrices, and return-to-risk measures.

• Generates up to 15 portfolios along a user-defined risk spectrum.

• Provides visual representations of the Efficient Frontier, Tangency Portfolio, and Minimum Variance Portfolio.

​

4. Mathematical Optimization Engine

• Performs multivariate optimization to align portfolios with specified risk targets.

• Allocates discrete stock quantities, applying rounding rules and sector-level constraints.

• Recalculates portfolio risk dynamically when allocations or constraints are modified.

​

5. Portfolio Controls and Constraints

• Constraints Module: Supports weight limits, sector caps, inclusion/exclusion rules, stock-type and various additional restrictions.

• Database Enhancement Engine: Filters and refines the investable universe prior to optimization.

​

Machine-Learning Components

1. Data Foundations

• Utilizes historical price data alongside technical and fundamental inputs.

• Employs walk-forward training procedures dating back to 2011 to simulate out-of-sample behavior.

2. Dual Random Forest Framework

• Random Forest Classifier: Estimates the probability of relative market outperformance.

• Random Forest Regressor: Estimates expected return magnitude.

​

The two models are used in combination to capture non-linear relationships while reducing sensitivity to overfitting.

​

3. Stock Selection Logic

• The classifier ranks stocks by likelihood of outperformance.

• The regressor ranks stocks by expected return.

• Outputs from both models are combined to form a candidate set designed to balance conviction and diversification.

​

4. Model Maintenance

• Models are retrained on a monthly schedule as new market data becomes available.

• This process is intended to maintain relevance under changing market conditions.

​

Operational Characteristics

• Human-in-the-Loop Control: Optimization outputs are generated algorithmically but remain fully reviewable and adjustable by the user.

• Explainability: Portfolio allocations, constraints, and risk metrics are explicitly calculated and traceable.

• Scalability: The system architecture allows extension to additional exchanges and asset classes, subject to data availability.

​

​