Ldahocentralcu · Learning Program
Quantitative Analysis and Machine Learning in Practice
A structured program for analysts, researchers, and engineers who want to move from spreadsheet reasoning to model-driven thinking. Delivered remotely, designed for learners across all time zones in the US.
Program structure
Six modules, one coherent arc
The curriculum runs 18 weeks and is organized so each module builds directly on the one before it - no disconnected side topics, no filler.
The first two modules establish the mathematical and statistical foundation: linear algebra, probability distributions, and hypothesis testing. Modules three and four move into supervised and unsupervised learning with Python-based labs using real financial and behavioral datasets. The final two modules address model evaluation, interpretability, and practical deployment considerations that practitioners actually encounter. Every lab session ships with annotated notebooks so learners can revisit the reasoning, not just the output.
Weeks 1–3
Mathematical foundations
Linear algebra, matrix operations, and the probability theory that underlies most ML algorithms.
Weeks 4–6
Statistical inference
Estimation, confidence intervals, A/B testing design, and when to distrust a p-value.
Weeks 7–10
Supervised learning
Regression, classification, tree methods, and ensemble approaches with hands-on Python labs.
Weeks 11–13
Unsupervised methods
Clustering, dimensionality reduction, and anomaly detection on real-world datasets.
Weeks 14–16
Model evaluation
Cross-validation, leakage, fairness metrics, and interpretability tools including SHAP values.
Weeks 17–18
Deployment and review
Packaging models, monitoring drift, and a capstone project reviewed by program instructors.
How it works
Designed around remote learners
The program was built from the ground up for asynchronous access - recorded lectures, downloadable notebooks, and forums that stay active between cohorts.
Each week releases 3–4 hours of recorded lectures segmented into 12–20 minute clips, so you can work through material between other commitments without losing the thread. Live sessions run four times per month and are recorded in full for those in different time zones. Discussion forums are moderated by instructors, not automated - questions get real answers, usually within 24 hours on weekdays. The capstone project in the final two weeks asks you to take a dataset you choose, build a defensible model, and document your reasoning at each decision point. It is not graded on outcome - it is reviewed on process.
The part that stuck with me was the section on model evaluation. I had been misreading cross-validation results for two years before this program pointed out exactly where my reasoning was off.
- Teodora Vanhanen, data analyst, Portland OR
18 weeks
Total program length
~5 hrs/week
Recommended study time
Python 3.10+
Primary lab environment
All US regions
Remote access, no local requirement
Quantitative reasoning
Building the habit of expressing problems numerically before reaching for a model. Covers measurement, estimation, and data quality assessment.
Feature engineering
Encoding domain knowledge into model inputs. Covers transformations, interaction terms, and what gets lost when you skip this step entirely.
Time-series analysis
Stationarity, autocorrelation, ARIMA, and where classical time-series methods interact - or conflict - with ML-based approaches.