I take a business problem, find the data problem underneath, and build the system that moves the metric.
I lead the team that ships it. Architecture first, then delivery, then production — data engineering through to agents that touch a system of record.
Client data is confidential, so I rebuild each system on public data and publish the code, the notebooks and the write-up. You never have to take my word for it.
What I have shipped, by business function and stack layer
One square per layer a system covers. The empty squares are honest — a full grid would be a lie.
all industries
| Data eng | Analytics | ML | LLM / RAG | Agentic | |
|---|---|---|---|---|---|
| Marketing | 2 | ||||
| Sales | |||||
| Customer | 2 | 2 | |||
| Operations | 2 | 2 | |||
| Supply chain | 2 | 2 | |||
| Finance | |||||
| Risk | |||||
| Pricing | |||||
| People / HR | |||||
| Energy / ESG |
Select a cell to see what's in it.
Grouped by metric family
Anonymised, relative. No client names, no absolute figures.
From the open rebuilds
Every system produces a teaching series. Full method, full code, nothing withheld.
Uplift modelling and incrementality
Why a churn score sends money to the wrong customers, and what to model instead. Builds to a working treatment-assignment loop with a permanent holdout.
Agents that touch production
Evals, human gates, and the failure modes that only appear once an agent can write to a system of record.
Forecasting for people who own inventory
Accuracy is not the goal. This series ends at a working-capital number, and treats forecast error as an input to an inventory policy.
I sit between the people who have the problem and the systems that solve it.
Someone in operations describes a mess in their own words. My job is to hear the data problem underneath, design the architecture, and lead the build until a number moves.
More about how I work →- Writing part 4 of the agents series — the trace chapter, which keeps growing.
- Rebuilding the fraud case-preparation agent on IEEE-CIS so the anchoring experiment is reproducible.
- Reading about constrained optimisation for price ladders. The pricing rebuild needs a better solver than the one I shipped.