Build Coff
Build Coff Interpretable ML for Financial Decision-Making
+27 62 307 5492 support@buildcoff.com

Build Coff - Woodstock, Cape Town

Machine learning explained for finance

Build Coff builds educational programmes that close the gap between statistical modelling and the decisions finance professionals make every day. Our focus is interpretable machine learning - understanding not just what a model predicts, but why.

Financial data analysis workspace with screens showing model outputs

What we actually teach

Finance teams regularly use models they cannot fully explain to auditors, regulators, or clients. That is a practical problem - one that interpretability tools like SHAP values, LIME, and decision-tree surrogates are designed to address.

Our programmes work through real credit scoring scenarios, fraud detection pipelines, and portfolio risk assessments. Each module builds from the mechanism of a model to the explanation layer, so learners can trace a prediction back to its inputs.

Business Decision
Model Explanation
Interpretability Layer
Feature Attribution
Raw Model

Interpretability stack - from model output to accountable decision

4
core modules covering credit, fraud, risk, and compliance
6+
years refining content with finance and data science practitioners
Lecture session on model explainability with SHAP visualisations
Interpretable ML in Finance
SHAP attribution
Credit scoring
LIME local fit
Audit trails
Fraud detection
Bias testing
Risk models
Regs SR 11-7
Team reviewing interpretable model outputs in a financial context

People behind the programmes

Each contributor brings a background split between working in financial services and applying machine learning in production - not just in research settings.

Nadia Botha

Curriculum Lead

Spent eight years in retail banking before moving into data science education. Structures every module around the questions a risk officer would actually ask during a model review.

Sipho Khumalo

ML Methods

Worked on fraud detection pipelines at two South African banks before joining Texnova. Focuses on making gradient boosting and neural network explanations legible to non-specialist audiences.

Riana Lötter

Regulatory Content

Covers the compliance side - SR 11-7 model risk guidance, IFRS 9 expected credit loss, and how explainability requirements are evolving across jurisdictions relevant to South African financial institutions.

Questions about programme content or institutional access? Reach the team directly at support@buildcoff.com

Get in touch