AI that can be trusted
Practical approaches to combining LLMs with deterministic systems, explicit rules, structured data and testable behaviour.
Simon M. J. Garrett
I work with machine learning, AI, graphs and messy real-world data.
Particularly where the evidence is incomplete, uncertain or difficult to connect. I am a data scientist and software engineer with a long-standing background in machine learning, interested in systems that help people reason well when the available evidence is fragmented, ambiguous or imperfect.
Selected themes
Fragmented evidence
Extracting useful conclusions when information is incomplete, inconsistent, uncertain or distributed across many sources.
Connected data
Knowledge graphs, entity resolution and related techniques for finding structure and relationships within complex data.
Applied machine learning
Using ML where it genuinely improves a system rather than adding it because it is fashionable.
Selected writing
Evergreen essays and notes will appear here when they are ready.
No public writing has been published on this site yet.