Credit risk modelling
Loss forecasting, IFRS 9 and CECL provisioning, PD models and stress testing, from development through governance and senior review.
- NatWest UK Retail
- JPMorgan Chase ICB Risk
Data engineering · Credit risk · Statistics
Stephen Ahiabah
I build the models banks use to forecast credit losses, and the data those models depend on.
Nine years at NatWest and JPMorgan Chase, across stress testing, IFRS 9 and CECL model development and, now, data engineering. Outside work I write CannonIQ, an Arsenal analytics newsletter read by 500+ people.
Simulated years0
Expected loss–
1 in 100 year loss–
Buffer above expected–
Illustrative one-factor (Vasicek) portfolio model. Losses in £k over one year. Each dot is one simulated year. Red bars are the worst 1% of years, the tail that stress testing and capital planning are built around.
What I work on
Three areas I keep coming back to
Most of my work sits where a model, a dataset and a business decision have to agree.
Data engineering
Pipelines, lineage and data quality controls that make critical data traceable and ready for analytics and AI.
- PySpark, SQL, Python
- BCBS 239 lineage
Statistics in public
Independent projects and writing where I test new methods, from Bayesian updating to Monte Carlo, and explain them in plain English.
- CannonIQ
- Pitch IQ
Track record
Selected results from nine years in banking
Numbers from my work at NatWest and JPMorgan Chase. The details are on the experience page.
| £200bn | UK retail portfolio covered by the stress testing and loss forecasting I co-led | NatWest |
| 60 → 10 min | IFRS 9 staging runtime after I directed its rebuild from SAS to Python | NatWest |
| 2 FTE | Recurring manual work removed by a roll-rate model and automated forecasting | JPMorgan Chase |
| 7+ | Inherited model risk issues cleared in my first year as lead model developer | JPMorgan Chase |
| 100+ | Colleagues moved onto SageMaker, CI/CD and GitHub across ICB Risk | JPMorgan Chase |
Writing · CannonIQ
Statistics in public, for 500+ readers
CannonIQ is my Arsenal publication on Substack. It's where I test new methods and explain them in plain English.
Model notes · 2026
Goals Are Rare
Poisson, weighted rolling averages, Bayesian updating and Monte Carlo, explained for football people.
↗ Expected points · 2026The Table Lies to You (A Bit)
Why a probabilistic view of team strength tells us something the league table can't.
↗ Recruitment · 2025Finding Arsenal's "Win Now" Forward
A Random Forest model for projecting goal contributions and testing shortlists.
↗ Long-form · 2026Same Old Arsenal. Same Old Lie.
A personal essay on fandom, narratives and what statistics can't settle.
↗