29 July 2026 · 8 min read
The 47th CV.
What building an AI agent for the job market taught me about volume, precision, and standing on someone else's shoulders. Notes on forking Santiago Fernández de Valderrama's career-ops into applyd, tuning it for the UK market, and drawing the boundary between deterministic code and language-model calls.
3 July 2026 · 5 min read
Dashboards inform, agents act.
Putting a Model Context Protocol endpoint on a churn model: what changes when an LLM agent can query survival hazards, SHAP drivers, and causal treatment effects directly, and why the data layer underneath decides whether the agent can be trusted.
More as I publish. Current activity also lives on LinkedIn.