AI Perspectives: How Humans Adapt and Language Models Work

Two talks at ShopFully's Milan office, hosted by PyData Milano. Raffaele Bongo, Senior AI Engineer at ShopFully, argues that AI agents promoted everyone from doing tasks to supervising the agents that do them, and that overseeing agents is now the most mentally tiring way to work with AI. His case is for a second promotion, from supervisor to orchestrator, and for designing agent systems a team can genuinely own. Luca Mattiazzi, AI Engineer at Pillar, takes garden-path sentences such as "The old man the boat" and uses mechanistic interpretability to trace how Gemma 3 models at 1B, 4B, 12B and 27B handle syntactic reanalysis, and how that shifts with scale. Doors 18:30, talks from 19:00, aperitivo from 20:30. Free, but RSVP is required and places were nearly gone at the time of writing. Why go: one of the few Milan rooms where applied agent engineering and interpretability research share a bill. Best for ML engineers and data scientists who want depth over a product demo.