Est. 2022Explainable ML
for banking professionals,
taught by practitioners
Domain runs focused online seminars on explainable machine learning for the financial sector. The material is built around real regulatory and risk management scenarios — not academic abstractions.
Banks need model explainability, not just accuracy
Regulators across Canada and internationally now expect financial institutions to demonstrate that their credit, fraud, and risk models can be audited and explained in plain language. Most ML courses skip this entirely.
Domain seminars address the gap between what data science teams build and what compliance, audit, and executive stakeholders actually need to see. Participants work through real bank scenarios — loan approvals, fraud flags, capital reserve models — and learn how SHAP, LIME, and monotonic constraints are applied in practice.
- Covers OSFI and Basel III explainability expectations directly
- Live discussion format — participants challenge assumptions with peers
- Structured case studies with anonymised real bank data examples
- No generic intro-to-ML content — prerequisites are stated clearly upfront
Multi-layer learning that fits how analysts actually work
Each seminar is built around three connected layers: the technical method, its regulatory context, and the communication challenge of explaining it upward to non-technical decision-makers.
Participants receive pre-reading material before each session — typically a short bank regulatory document and one research paper. Sessions run 90 minutes with structured discussion built in from the first 20 minutes.
Layer 1 — Technical method
How SHAP values, partial dependence plots, and counterfactual explanations work in gradient boosting and neural network models used in credit scoring.
Key tool focus
SHAP, LIME, InterpretML — applied to tabular financial data
Layer 2 — Regulatory fit
How explainability outputs map to model risk management frameworks — SR 11-7, OSFI E-23, and internal model validation requirements.
Document coverage
Regulatory guidance notes provided per session, not as a separate reading list
Layer 3 — Stakeholder communication
Translating technical outputs into language risk officers and board committees can act on — without losing precision.
Practice format
Peer critique rounds and structured Q&A simulations
Small cohorts
Groups of 4–6 mean every participant gets time to work through their specific use case, not just listen.
Async follow-up
After each session, a written summary with annotated code snippets is shared so participants can reference material without rewatching recordings.
Iterative curriculum
Topics are updated each quarter to reflect current regulatory developments — participants from prior cohorts flag what has changed in their institutions.
Small team, direct access to instructors
Domain is built around a small group of practitioners with backgrounds in quantitative risk, model validation, and applied ML at financial institutions. Instructors have held roles at banks and credit unions before moving into education.
Participants communicate with instructors directly — there are no teaching assistants intermediating the technical discussions. Questions raised between sessions are answered by the person who ran the session.
Petra Vanlith
Lead Instructor, Model Risk
