New models are bigger and more capable than ever, and serving them fast, reliably, and affordably is one of the hardest problems in AI engineering.
Join Baseten and Rootly in Toronto for an evening with engineers who run LLM inference in production every day. Our panel will cover speculative decoding and the other techniques teams use to make models faster, along with what actually matters once real customer traffic hits.
Topics:
-
How speculative decoding works, and when it’s worth it
-
What’s changed now that newer models ship with strong speculators out of the box
-
Tuning inference performance for specific customer workloads
-
Keeping LLM systems reliable in production when things break
Panelists
-
Sylvain Kalache, Head of AI Labs, Rootly
-
Daniel Buades Marcos, Engineering Manager, Forward Deployed Engineering, Baseten
-
More speakers to be announced
Schedule6:00–6:30 PM: Welcome drinks and snacks6:30–7:30 PM: Panel and Q&A7:30–8:00 PM: Wrap up and networking
📍 Rootly Toronto - 366 Adelaide St W, Unit 601🗓 Thursday, November 19th
