BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//cfp.devcon.org//devcon8//talk//PKTCGR
BEGIN:VTIMEZONE
TZID:IST
BEGIN:STANDARD
DTSTART:20000101T000000
RRULE:FREQ=YEARLY;BYMONTH=1
TZNAME:IST
TZOFFSETFROM:+0530
TZOFFSETTO:+0530
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-devcon8-PKTCGR@cfp.devcon.org
DTSTART;TZID=IST:20261105T110000
DTEND;TZID=IST:20261105T111500
DESCRIPTION:KServe can serve large language models on Kubernetes - but past
  "hello world\," it gets sharp. You face disaggregated prefill/decode\, fo
 ur kinds of parallelism\, autoscaling that behaves differently under KEDA 
 versus HPA\, a CPU profile that runs out of memory before your model loads
 \, and gateway wiring where one wrong setting silently breaks routing. Mis
 s a field and you don't get an error - you get a slow\, wrong\, or unreach
 able model.\n\nThis talk shows how the OpenEverest's KServe provider hides
  all of that behind a single\, technology-agnostic Instance resource. In t
 en lines of YAML\, you get an OpenAI-compatible endpoint. We'll look at th
 e real traps it saves you from: container args that silently drop your def
 aults\, a gateway backend that bypasses request routing\, and KV-cache set
 tings that refuse to start.
DTSTAMP:20261101T122957Z
LOCATION:Community Led Stage
SUMMARY:Prefill\, Decode\, and Despair: Disaggregated LLM Serving Made Bori
 ng - Sergey Pronin
URL:https://cfp.devcon.org/devcon8/talk/PKTCGR/
END:VEVENT
END:VCALENDAR