Chris Tyson

Senior AI Engineer at Elsewhen, working on RAG and agentic systems. Previously Head of Machine Learning at glass.ai, researching AI and ML approaches for building bespoke company datasets semi-autonomously. Before that, dashboarding at JPMorgan Chase.

Interested in AI safety, mechanistic interpretability, and computational social science - three angles on the same instinct, to understand systems (human, algorithmic, or their interaction) from the traces they leave. My MSc thesis at Exeter explored this for electorates, using 2019 UK general election Twitter data to infer constituency-level political attitudes. More recent reading has been around the interpretability of reasoning in language models, and the dynamics of long-context conversational harm.