Why it matters
Concentrated exposure to the inference-serving stack (SGLang, Crusoe GPUs, dstack orchestration) that determines how cheaply and quickly models actually run in production.
The tokenmaxxing angle
SGLang and Crusoe are core levers in inference cost — serving engine choice and GPU provisioning directly set the dollars-per-million-tokens number teams are trying to optimize.
From the organizers
12 confirmed speakers (incl. Yuwei An, Andrey Cheptsov, Steeve Morin, Dmitri Melikyan) each get a 5-minute lightning-talk slot starting 6pm; entry is 21+ with ID and capacity-capped, host-approval RSVP.