It is an interesting architecture. Flink for real time feature computation and Kafka to share training samples with clients for low end to end latency. Nothing

It is an interesting architecture. Flink for real time feature computation and Kafka to share training samples with clients for low end to end latency. Nothing on use of embeddings for similarity search, though. I assume they are using FAIS?