Where machine learning meets portfolio decisions
Automated investment systems have been generating noise for years. The ones that hold up are built on statistical discipline, not optimism. Waeolu's workshops teach you to build the ones that hold up.
See the program
across 3 skill tiers
Three tiers, one trajectory
Each tier builds on the previous. Skipping ahead is possible if you can demonstrate the prerequisite skills — the intake assessment takes about 40 minutes.
| What you get | Foundation | Applied | Advanced |
|---|---|---|---|
| Live workshop sessions Instructor-led, real-time problem solving | 8 sessions | 12 sessions | 16 sessions |
| Backtesting lab access Historical data environments for strategy testing | 30 days | 90 days | 180 days |
| Code review with instructor 1-on-1 feedback on your model implementations | |||
| Portfolio risk simulation Stress-test your allocation models against scenario data | |||
| Capstone project mentorship Guided end-to-end system build with feedback cycles | |||
| Peer collaboration tools Shared workspaces and async review threads | |||
| Recorded session library Full replay access for all attended sessions | 60 days | Unlimited | Unlimited |
| Pricing per participant | CA$890 per cohort | CA$1,490 per cohort | CA$2,200 per cohort |
Signal extraction and feature engineering
Raw financial data rarely arrives in a form that models can use directly. Sessions cover cleaning, normalizing, and constructing features from price series, volume data, and macroeconomic indicators.
- Rolling window statistics
- Lag feature construction
- Correlation filtering
- Regime detection inputs
- Alternative data parsing
- Feature importance ranking
Model selection and validation under real constraints
Choosing a model without understanding its failure modes is how most automated systems fall apart. Workshops cover walk-forward validation, out-of-sample testing, and the specific pitfalls of financial time series.
Backtesting methodology and performance attribution
A backtest that flatters is worse than no backtest at all. Participants work through common sources of look-ahead bias, survivorship bias, and overfitting — then learn to read performance reports critically rather than optimistically.
- Sharpe and Sortino ratios
- Maximum drawdown analysis
- Transaction cost modelling
- Slippage estimation
Portfolio construction and allocation logic
Signal quality only matters if the allocation logic doesn't cancel it out. Sessions cover mean-variance optimization, risk parity, and constraint-based allocation — with hands-on exercises using real equity and fixed income data.
Risk management systems and position sizing
Automated systems that ignore risk management tend to work until they don't, then fail badly. Participants build stop-loss logic, volatility-scaled sizing, and drawdown circuit breakers directly into their strategy code.
- Kelly criterion variants
- ATR-based stop placement
- Volatility targeting
- Correlation-adjusted sizing
- Drawdown circuit breakers
- Exposure limit enforcement
Deployment, monitoring, and system maintenance
Getting a model to production is a different skill than building one. Advanced tier participants deploy their systems to a sandboxed brokerage API, set up monitoring dashboards, and work through a simulated model degradation scenario.