louis grove · senior user researcher
I study how people make decisions inside complex, evolving systems — first biological, then social, now AI-native.
Mixed-methods researcher with 7+ years of practice. I build research tooling I want to use, with safety, evaluation, and observability as load-bearing design constraints: an evaluation-and-curation system for an internal library of agentic workflows; a perimeter-deployed Slack helper with human-in-the-loop fallback, least-privilege tool access, and full interaction telemetry; instrumentation for AI-augmented research workflows across the company. Quantitative rigour paired with deep qualitative methods, and a working obsession with the feedback loops that make systems improve themselves — and how those same systems can mislead or be abused, intentionally or otherwise.
selected projects
Smart Skills Platform
shipped · may 2026An evaluation and curation system for an internal library of agentic workflows. Failed prompts cluster into evidence; curators promote on signal; new development is directed at the gaps the runtime surfaces. Built in 3 days at an internal hackathon.
User Research Slack Helper
in productionPerimeter-deployed Slack helper with least-privilege tool access, human-in-the-loop fallback for out-of-scope queries, and full interaction telemetry. Answers colleague questions from internal context; routes clarifying questions to the user research team.
Inscryption Agent
in progressPersonal project building an RL agent for the deckbuilder Inscryption (Kaycee's Mod). Validated Rust simulator complete and surfacing actionable play patterns; warm-start training data run done; agent training next. Most transferable artefact so far: a framework for parallel agentic workflows.
Intersectional Decision-Support Tool
concept · scopingConcept: decision-support tool grounded in intersectional analysis (Crenshaw, 1989; Hill Collins). Recognises that interlocking systems of power compound multiplicatively to produce qualitatively distinct user contexts; designs at those intersections to produce disproportionate inclusive uplift while surfacing risks of harm flat user models obscure.