File
Model-risk lite
Purpose, data, limits, monitoring, challenger, owners.
Financial Services AI
Credit, fraud, support and ops in a regulated firm: model risk, explainability that is real, and audit trails that survive a second line.
Created by Baljeet Dogra
Course Objectives
- Separate marketing “explainability” from what a reviewer will accept.
- Design an audit trail for a generative or scoring system.
- Name fair-lending and conduct risks in the design, not the appendix.
- Walk a second-line reviewer through the file in twenty minutes.
What you will produce
Build
Ops or fraud slice
A narrow system with traces and a documented override.
4 weeks curriculum
Expand a part for the syllabus. Content stays searchable when closed.
01 The firm Week 1
Three lines of defence. Where AI actually lands: ops, fraud, support, not a trading fantasy.
- 3LoD
- Use cases
- Conduct
- Data
02 Model risk Week 2
Purpose, challengers, monitoring, what “material” means.
- MRM
- Challengers
- Limits
- Monitoring
03 Explain and audit Week 3
Traces, reasons, human override, fairness metrics that are specified.
- Traces
- Overrides
- Fairness
- Logs
04 Review Week 4
Second-line walkthrough. Capstone file plus slice.
- Review
- Gaps
- Residual risk
- Capstone
Who this is for
Engineers in banks and fintechs
You ship. Second line will ask questions you should have designed for.
Risk and model-risk staff
You need generative-system literacy, not only scorecards.
Prerequisites
- Comfortable with APIs and data
- You can write for a non-engineering reviewer
Not a fit if
- High-frequency trading courses
- People who want to skip audit because “it’s just an LLM”
Related: AI Governance & Compliance · AI Testing & QA
Questions
FCA or PRA?
UK-aware language. You map to the firm you have; we do not pretend one file fits all licences.
Will we predict markets?
No. Ops, fraud and support are where most engineers are actually asked to ship.
Ready to start?
Four weeks to a file a second line can review. Create an account to enrol.
Enrol now