Hands-on explorations. I don't only write strategy — I build and test.
Role-based multi-agent systems that plan, use tools and coordinate on multi-step tasks.
Prompt engineering, evaluation and AI-assisted workflows built hands-on.
Retrieval, embeddings and grounding that keep AI answers cited and traceable.
SQL, Python and R turned into tools — from prioritisation models to dashboards.
Pipelines wiring Figma to AI models to ship coded prototypes fast.
A shipped RAG product — grounded, source-cited answers from messy source material.