The IAS – Institute for Advanced Study programme, developed by the Istituto Italiano di Intelligenza Artificiale (AI4I) in collaboration with Fondazione Compagnia di San Paolo, continues its seminar series with a new appointment focused on one of the most critical challenges in artificial intelligence: bridging the gap between experimental models and real-world deployment.
The IAS – Institute for Advanced Study programme, developed by the Istituto Italiano di Intelligenza Artificiale (AI4I) in collaboration with Fondazione Compagnia di San Paolo, continues its seminar series with a new appointment focused on one of the most critical challenges in artificial intelligence: bridging the gap between experimental models and real-world deployment.
Following the two major public events held in March – the conference by Konstantin Novoselov and the Lectio Magistralis 2026 delivered by Jürgen Schmidhuber – the IAS programme moves forward with a more technical and application-oriented focus, confirming its structure alternating strategic reflection and operational insight.
On April 28, 2026, IAS Seminar #04/2026 will take place at OGR Torino (Sala Duomo), from 4:00 to 6:00 pm CEST, under the title “From Benchmarks to Reality: Lessons from Deployed AI Systems”.
The seminar will feature Andrea Ferretti, AI Research Engineer at Bending Spoons, currently working on language models for Evernote and previously involved in computer vision systems for the Remini app.
The talk will address the challenges that emerge when artificial intelligence models are deployed in real-world environments: the gap between performance in controlled settings and behaviour in production, the impact of distribution shift, unpredictable user inputs, latency and cost constraints, as well as the need to meet user expectations while ensuring safety and moderation.
Through two concrete case studies – a feature in the Remini app and the system powering Evernote’s AI Assistant – the seminar will provide an in-depth perspective on the dynamics of AI industrialization, highlighting trade-offs, iterative development processes and the solutions adopted to ensure robustness and reliability.
The IAS programme thus continues to position itself as a structured platform connecting advanced research with industrial application, in line with AI4I’s mission to integrate research, engineering development and technology transfer, and to support the effective adoption of artificial intelligence within production systems.
To learn more and register:
