I build trusted, cloud-native platforms for complex quantitative decision-making and regulated infrastructure.
I lead credit risk model platform delivery in financial services, working hands-on across Python, Azure, APIs, SDKs and model lifecycle controls. My focus is the engineering around the model: clear contracts, usable tools, repeatable execution and evidence people can inspect.
|
Critical model tests / day Up from 2–3 runs/day |
Modeller teams Hierarchical Python SDK patterns |
Concurrent users Self-service model interrogation |
Systems / Stack / Engineering practice / Activity / Background
|
01 / MODEL PLATFORMS
Execution, monitoring, impact assessment and scenario testing, with data lineage, audit evidence and controlled release paths. Self-service analytics and platform abstractions that let quantitative teams scale independently. |
02 / PYTHON APIs & SDKs
FastAPI services, Pydantic validation, typed contracts, SQLAlchemy data access and SQL optimisation. Layered SDKs, package and version governance across teams, and test-first delivery focused on observable behaviour. |
03 / Data & decision systems
Time-series and heterogeneous data pipelines, Monte Carlo simulation, numerical optimisation and PD/LGD calibration. Separate data, assumptions and model outputs so the decision logic remains visible.
Implementation detail — cloud services, delivery and governance
| Area | Tools & patterns |
|---|---|
| Azure | Container Apps, Functions, Azure SQL, Cosmos DB, Azure Machine Learning, Application Insights |
| Delivery | Azure DevOps, CI/CD, Docker, coverage gates, vulnerability scanning |
| Architecture | API-led systems, SDK layering, event-driven workflows, cloud-native deployment |
| Data & analytics | Time-series workflows, simulation, optimisation, pandas, NumPy |
| Governance | Model lifecycle controls, audit evidence, data lineage, controlled release patterns |
| AI-assisted work | Structured context, reusable agent workflows, code review support, documentation acceleration |
Boring in production. Expressive in development.
| Boundary | What I make explicit |
|---|---|
| Data → models | Validated inputs, typed contracts and traceable assumptions. |
| Models → workflows | Composable layers; separate model logic from orchestration and presentation. |
| Workflows → decisions | Observable execution, explainable outputs and evidence that can be reviewed. |
| Development → production | Behaviour-focused tests, automated delivery and controlled release paths. |
Currently exploring: AI-assisted engineering with structured context and repeatable workflows for review and documentation in controlled environments. Alongside this, I continue to focus on Python developer experience, Azure-first analytical platforms and regulated model governance.
Explore my public repositories →
I started in automotive engineering: powertrain optimisation, emissions modelling and predictive analytics. Working with physical systems, noisy time-series data, neural networks, simulation and numerical optimisation still shapes how I build financial technology: understand the system, measure what matters and make hidden complexity visible.
Delivery evidence — regulated platforms and engineering analytics
- Led credit risk model platform delivery across IFRS 9, IRB and FCA MCOB 11.6 contexts.
- Supported Big Four audit reviews with no findings within owned implementation and control evidence scope.
- Increased critical model testing throughput from 2–3 to 8 runs/day by removing infrastructure and workflow bottlenecks.
- Built self-service model interrogation tooling for 100+ concurrent users.
- Designed hierarchical Python SDK patterns used across 15+ modeller teams.
- Automated around 70% of recurring analytical workload in automotive predictive analytics.
- Designed data-processing pipelines for engineering datasets at 50M+ daily observations.
- Developed emissions simulation methodology later adopted as a Ford Motor Company standard.
Complex models. Clear interfaces. Decisions you can inspect.