François Chollet's Recent LinkedIn Posts

Jeremy Berman, former #1 on the ARC-AGI-Pub leaderboard, just open-sourced his solution as a template on Params: https://lnkd.in/gpCWHSBv It uses LLM-driven genetic programming, which has turned out to be way more powerful than anybody expected. You can book a consulting call with Jeremy via Params to talk about how the same approach could apply to your use case.

Trying something new: I signed up for Params, a service that lets you browse curated ML project templates from open-source developers and book calls directly with the authors. I'm starting with a flexible recommender systems template in Keras 3, that works with JAX, TF, and torch. If you're looking to add recommendations to your product, take a look at this codebase -- you should be able to make use of it easily. Here's the template link: https://lnkd.in/ggGQuN4i It covers both cases where you only have binary interaction data (e.g. listens of a song) and cases where you have user ratings. Unlike a matrix factorization approach, it lets you incorporate arbitrary user features and item features into the predictions (thanks, Keras FeatureSpace API!) You'll see there's a "book a call" button. It lets you book a consulting call with me about the template in a couple of clicks. We can talk about your use case, how to customize the template, how to add any feature you have in mind, etc. Here: https://lnkd.in/g7KtrC7n