Can agents improve my product retention? Drinks with PMs and ML leaders @ GTC

​You ran the experiment. The model looked great offline. And the metric barely moved.

​If you work on consumer products, ecommerce, or marketplace platforms, you've been here. Moving business metrics with ML is one of the most frustrating and least-discussed challenges in the field, and it's especially acute when the stakes are high: recommendation quality that doesn't lift conversion, pricing models that look clean in testing but flatten in production, search ranking changes that shift one metric while quietly hurting another, notification systems that optimise for opens but erode long-term retention.

​Join a small, invite-only dinner for Product Managers, ML engineers, and data leaders at GTC who are working on exactly these problems, in consumer internet, ecommerce, DTC, and marketplace businesses.

​We'll get into the real reasons it's so hard: proxy metrics that don't generalise, experiment designs that can't detect real effects, feedback loops that take months to close, organisational friction between product and data teams, and the gap between what a model optimises for and what the business actually cares about.

​No slides. No pitches. Just an honest conversation with people who've been in the same trenches.

​Spaces are very limited. If you're at GTC and want to spend an evening talking about the hard parts of applied ML in consumer and commerce products, we'd love to have you.

​Details Wed March 18 📍 Walking distance of San Jose Convention Center, TBD (confirmed with attendees closer to the time) 🍽️ Dinner, drinks on us

​Hosted by Pavo AI — built by the team behind Spotify's recommender systems and feed ranking, with 15+ patents and 100+ research papers across ML and data systems.