Machine Learning for Portfolio Optimisation
Structured seminars that take you through applied ML methods - from data preparation to live portfolio decisions. Built for practitioners who need working knowledge, not just theory.
What participants study across our seminar tracks
Each seminar is structured around a specific problem in portfolio management - not a broad survey. You leave with a method you can test against your own data, not a slide deck full of concepts.
Sessions run in small cohorts with structured peer discussion built into every module. The format has been refined since 2015 based on what actually helps practitioners apply ideas after the seminar ends.
Factor Model Construction
Participants build multi-factor models using regularised regression and feature selection. The seminar covers how to avoid data snooping and evaluate factor decay over time.
6 sessions · AppliedRisk-Aware Allocation
Covers covariance estimation under non-stationarity, shrinkage methods, and how ML-based forecasts feed into mean-variance and Black-Litterman frameworks without overfitting.
8 sessions · IntermediateRegime Detection
Hidden Markov Models and clustering approaches applied to identifying market regimes. Participants learn how regime labels affect position sizing and rebalancing frequency.
5 sessions · AdvancedAlternative Data Integration
Structured seminars on processing satellite data, earnings call transcripts, and web-scraped signals. Covers NLP pipelines and how to assess signal quality before committing capital.
7 sessions · AppliedBacktesting Without Bias
Walk-forward validation, combinatorial purged cross-validation, and transaction cost modelling. A practical seminar on the gap between backtest performance and live results.
4 sessions · CorePeer Discussion Cohorts
Each track includes structured peer sessions where participants present their own implementations, challenge assumptions, and get feedback from others working on similar problems.
Ongoing · All tracksEnrol and pick a track
Choose based on your current role and the specific portfolio problem you are trying to solve. Tracks are independent - no prerequisite sequence required.
Work through structured modules
Each session combines a focused lecture segment with a hands-on exercise using real market data. Python notebooks are provided for every module.
Participate in live discussion
Scheduled group sessions run weekly. Participants share implementations, ask questions, and critique each other's approaches in a structured format.
Apply to your own portfolio
The final module in each track focuses on adapting the methods to your specific asset universe, constraints, and data availability.
What a typical track covers
5 of 7 core topics completed in sample track
- Data sourcing and cleaning for financial time series
- Feature engineering and signal construction
- Model selection and hyperparameter tuning
- Portfolio construction with ML-derived inputs
- Validation and out-of-sample testing
- Live implementation and monitoring
- Peer review and final presentation
From participants across our cohorts
The backtesting module changed how I evaluate strategies. I had been making the same cross-validation mistake for two years without realising it. The peer sessions were where most of the real learning happened.
I came in sceptical that ML would be practical for the fund sizes I work with. The regime detection track gave me a framework I could actually apply. It took about three months of work after the seminar to get something running.
The covariance estimation content was the most technically dense thing I had studied since my postgrad. The instructors were direct about where the methods break down, which made it much more useful than textbook treatments.
Questions about which track fits your situation?
Reach out directly - support@quickprem.com or +27 21 935 0612. The team can walk you through the track options based on your background and goals.