About this program
Portfolio management has a lot of moving parts, and ML does not simplify them — it shifts where the complexity lives.
Reinforcement learning in allocation problems
Spend the first third of the program on RL fundamentals applied specifically to multi-asset allocation. Use OpenAI Gym environments built around equity and bond data to train agents that learn position sizing under simulated market conditions. The focus is on understanding what the agent is actually optimizing, not just running training loops.
Ensemble methods for signal aggregation
Single models break. Ensemble approaches — stacking gradient boosted trees with neural forecasters, for instance — tend to be more stable across different market regimes. Work through several architectures and measure how each one degrades when market conditions shift from the training period.
Execution quality and market impact
Automated systems that ignore execution quality often look good in backtests and disappoint in practice. Cover basic market microstructure: bid-ask spreads, order book depth, and how to model the cost of moving in and out of positions at scale.
Risk frameworks alongside ML
Covers Value at Risk estimation using ML, conditional drawdown constraints, and factor exposure monitoring. These are not replacements for traditional risk management but additions that can catch things rules-based systems miss.