The GSA Capital Portfolio Construction & Machine Learning interview is stage 4 of 5 in the GSA Capital interview process, after the Technical. Format: 45–60 min · with a researcher. A round on portfolio optimization and, increasingly, machine learning.
Preparing for this round? Practice the 8 questions mapped to the GSA Capital Portfolio Construction & Machine Learning, with worked solutions. A free account saves your progress.
What the Portfolio Construction & Machine Learning tests
Expect Markowitz optimization given an alpha signal and covariance matrix (objective, constraints, unconstrained closed form), covariance-estimation methods (sample vs EWMA vs PCA factor models vs graphical Lasso), and recent ML items: PCA components with near-zero R², tree feature importance / SHAP values, and neural-net initialization / batch-norm-and-dropout at test time.
QuantVault maps 8 practice problems to this round. By topic, the biggest groups are machine learning (4), statistics (2), optimization (1), regression (1); by difficulty, 0 easy, 7 medium and 1 hard. 8 of them were reported by candidates who interviewed at GSA Capital, most recently in Jul 2026.
- This content belongs to the second round, up to about four hours of deeper, team-specific interviews.
How to prepare for the Portfolio Construction & Machine Learning
- Set up mean-variance cleanly. Maximize w'μ − (γ/2)w'Σw; unconstrained gives w* ∝ (1/γ)Σ⁻¹μ; know the sum-to-one Lagrange version.
- Have a covariance-estimation menu. Sample covariance breaks when N ≫ T; be ready to compare EWMA, PCA factor models, and graphical Lasso (precision-matrix MLE).
- Connect ML to linearity. Low R² on top PCA components often just means the signal is nonlinear; know that a univariate SHAP value reduces to the regression beta.
Further reading: Behavioral questions at trading firms · Fermi estimation questions · How quant superdays work · Market-making game interview guide.
Question types in this round
The 4 largest question types in the GSA Capital Portfolio Construction & Machine Learning practice set; across them the most common sub-topics are feature engineering, model selection, regularization, normal distribution. The questions themselves, with worked solutions, are in the round practice.
Machine Learning4 in this round
Statistics2 in this round
Optimization1 in this round
Regression1 in this round
Prepare with QuantVault
Compare with similar firms
Frequently asked questions
What is the GSA Capital Portfolio Construction & Machine Learning interview?
It is stage 4 of 5 in the GSA Capital interview process, after the Technical. Format: 45–60 min · with a researcher. A round on portfolio optimization and, increasingly, machine learning.
How long is the GSA Capital Portfolio Construction & Machine Learning interview?
Candidates report: 45–60 min · with a researcher.
What questions come up in the GSA Capital Portfolio Construction & Machine Learning interview?
The practice set QuantVault maps to this round centers on machine learning, statistics, optimization.
How should I prepare for the GSA Capital Portfolio Construction & Machine Learning interview?
Set up mean-variance cleanly. Maximize w'μ − (γ/2)w'Σw; unconstrained gives w* ∝ (1/γ)Σ⁻¹μ; know the sum-to-one Lagrange version. Have a covariance-estimation menu. Sample covariance breaks when N ≫ T; be ready to compare EWMA, PCA factor models, and graphical Lasso (precision-matrix MLE).
What comes after the Portfolio Construction & Machine Learning?
The Final (stage 5 of 5). Format: Conversational · with senior researchers.
Practice the GSA Capital Portfolio Construction & Machine Learning round
Work the 8 mapped practice problems with progress tracking, then move to the next stage of the GSA Capital funnel.