Work
Real projects, real metrics. Three engagements that show what AI and agile delivery looks like in practice — from enterprise transformation to startup MVP.
Major European energy company, 6 business units, 40+ people. SAFe transformation from waterfall without stopping delivery, then the AI layer: a two-agent pipeline replacing manual backlog writing, and a phased program embedding AI agents into the development cycle across 23 microservices. The transformation work and the AI work reinforced each other.
Major logistics company. Jira on-premise → cloud migration for 100+ people, multiple legacy systems, unrealistic timelines pushed by management. Used flow metrics to kill the bad timeline before it killed the project. Zero scope creep, zero post-production firefighting.
Built from zero, as architect and builder, not as a PM writing specs. Designed the retrieval architecture (AREX: adaptive RAG with conceptual graphs and multi-thread retrieval), configured the agentic workflows on n8n, integrated Claude API with governance constraints, tested the endpoints. The product lets a small company query its own internal documents without the AI inventing answers. Working MVP with real users in 2 months. 30+ structured interviews with Italian SME owners shaped what got built.