Machine Learning
for Investment
Decisions
A structured program built around real portfolio problems - not abstract theory. You'll work through data pipelines, model selection, and risk evaluation using tools practitioners actually use.
Six modules, two learning paths - [choose your pace]
The program runs as both group cohorts and individual sessions. Group cohorts give you peer review, scheduled live calls, and shared datasets. Individual sessions let you set your own timeline and bring your own investment problem.
Either way, you move through the same six modules. The difference is in the rhythm, not the depth.
-
1
Data sourcing and cleaning
Working with financial time-series data - equities, ETFs, macro indicators - using pandas and yfinance. Handling missing values, survivorship bias, and look-ahead leakage.
-
2
Feature engineering for markets
Building meaningful signals from raw price and volume data. Rolling statistics, cross-asset correlations, and regime indicators that hold up in walk-forward tests.
-
3
Model selection and validation
Comparing gradient boosting, LSTM sequences, and ensemble approaches on the same dataset. Emphasis on avoiding overfitting through proper cross-validation design.
-
4
Portfolio construction and sizing
Translating model outputs into allocation decisions. Kelly criterion, volatility targeting, and constraint-based optimisation with realistic transaction cost assumptions.
-
5
Risk monitoring in live conditions
Setting up drawdown limits, model degradation alerts, and rebalancing triggers. Understanding when a model has stopped working versus when the market has changed.
-
6
Capstone project review
Each participant presents a complete strategy - from data to backtest to risk framework - reviewed by an instructor with written feedback and a final Q&A session.
Group Cohort
8–12 participants per cohort. Weekly live sessions on Tuesdays, shared Slack workspace, and peer-reviewed assignments. Cohorts run over 10 weeks.
CollaborativeIndividual Sessions
One-to-one sessions scheduled around your availability. You set the pace, bring your own data if you have it, and get direct feedback without waiting for a cohort schedule.
Personalised
Practical skills, not just
"model awareness"
Most participants come in knowing Python basics and having read about ML - but without a clear picture of where models break down in financial contexts. The program is designed around that specific gap.
By the end, you'll have built and stress-tested at least one complete strategy pipeline. That's the baseline, not the ceiling.
- Ability to identify and correct data leakage in financial datasets
- Working knowledge of scikit-learn, LightGBM, and basic PyTorch for sequence models
- A documented backtest with realistic slippage and cost assumptions
- A risk framework you can adapt to new strategies going forward
The instructors
Prerequisites
Python at an intermediate level - you should be comfortable with loops, functions, and basic data manipulation. No prior ML or finance background required, though familiarity with either helps.
Ask About Eligibility