Not a chronological CV, but a map of recurring problem types: uncertain evidence, connected data, applied AI and analytical systems that need to be understood as well as used.
Designing analytical systems where statistical models need to earn their place: improving decisions, reducing ambiguity or making complex evidence easier to work with.
Finding useful structure in data that is spread across sources, schemas and vocabularies, with attention to provenance and the limits of what the graph can really say.
Working out when fragmented records refer to the same underlying person, organisation, place or event without pretending the answer is always certain.
Combining language models with explicit rules, structured data, tests and constraints so that AI-enabled systems behave like engineered software rather than demos.
Reasoning across incomplete, inconsistent and distributed evidence while preserving uncertainty instead of smoothing it away too early.
Building tools for people who need to inspect, challenge and reuse analytical conclusions, not just consume a dashboard or a model score.
A small set of public artefacts can be added here over time without turning the page into a project archive.
No public projects are listed yet.