AI is taking coding by storm, but adoption for data-engineering tasks is lagging. This is not surprising: data workloads are ops-heavy, stateful, and cloud-dependent, whereas much software engineering can be approximated by a local codebase. What would it take for agents to be lakehouse-ready for production data? Drawing on the data-management literature and the MVCC model, we argue that the agentic lakehouse is the concurrent lakehouse, i.e., a data system designed for dozens of agents to collaborate safely. We demonstrate a self-repairing pipeline by running frontier models on Bauplan and close with a preview of our latest AI-for-systems research.
The short talk will be followed by an Ask-Me-Anything session where you will be able to ask Jacopo anything related to data, career advice or whatever (did you know Jacopo worked with Olimpia Milano at the beginning of his career?).
Jacopo Tagliabue, Co-founder @ BauplanJacopo Tagliabue is the co-founder of Bauplan, a data-infrastructure company based in SF. Previously, he co-founded Tooso (acquired by TSX:CVO), led Coveo's AI through IPO and launched Coveo Labs, producing open-source libraries, models, and datasets widely adopted by industry and academia. His research spans information retrieval, data management, and computer systems, with collaborations (among others) at Netflix, NVIDIA, Stanford, the University of Wisconsin–Madison. While building his new startup, he moonlights as Professor of ML Systems at NYU, which is only notable because it is the only job he ever had that his parents understand.
00:00 - Intro