AI search, discovery, and recommendations from one index: a live demo on real media catalogues
"A live demo of practical AI in production: hybrid search, combining full-text matching with semantic vector retrieval, running on a real media catalogue.
The demo uses RMC BFM Play's catalogue (EPG and media library), where Meilisearch search runs in production. It covers the queries media platforms see every day: misspelled titles, partial plot descriptions, and multilingual input, all resolved in milliseconds against real-world metadata.
The same index can do more than search. The demo shows recommendations built directly on it: similar-content suggestions, natural-language questions, and personalisation from a platform's own first-party data, with no separate recommendation engine and no data leaving the stack.
A second deployment shows the engine on a very different workload. Sky News uses Meilisearch in its internal article catalogue, enabling journalists to retrieve content quickly across an archive where relevance shifts by the hour.
For CTOs and engineers who own search or discovery infrastructure: real queries and real latency numbers, not slides. Questions welcome at the Meilisearch stand after the session. "