BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.devconf.info//devconf-us-2026//speaker//PMVQQR
BEGIN:VTIMEZONE
TZID:EST
BEGIN:STANDARD
DTSTART:20001029T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10;UNTIL=20061029T070000Z
TZNAME:EST
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
END:STANDARD
BEGIN:STANDARD
DTSTART:20071104T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11
TZNAME:EST
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000402T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=4;UNTIL=20060402T080000Z
TZNAME:EDT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:20070311T030000
RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3
TZNAME:EDT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-devconf-us-2026-7CPYUJ@pretalx.devconf.info
DTSTART;TZID=EST:20260924T100000
DTEND;TZID=EST:20260924T103500
DESCRIPTION:Reinforcement learning for LLMs still looks like something only
  big labs can afford: expensive GPUs\, complex training stacks\, fragile r
 ecipes\, and weeks of trial and error. But the ecosystem has quietly chang
 ed. With LoRA-based training\, TRL\, Unsloth\, and adapter-aware serving i
 n systems like vLLM and llama.cpp\, it is now realistic for ordinary engin
 eers to run useful GRPO experiments and serve multiple task-specific adapt
 ers on shared infrastructure.\n\nThis talk is for engineers who want to mo
 ve beyond prompting and try practical RL post-training without building a 
 research lab. We’ll walk through a small GRPO training recipe using LoRA
 \, explain the 5–6 hyperparameters that matter most\, and show how those
  choices affect training behavior in practice.\n\nAttendees will leave wit
 h a runnable notebook\, a mental model for configuring GRPO runs\, and a p
 ractical path from “trained LoRA adapter” to “servable model variant
 ” on existing inference infrastructure.
DTSTAMP:20260727T165239Z
LOCATION:Ladd Room (Capacity 170)
SUMMARY:The Practical Engineer’s Guide to RL Post-training - Rohan Awhad
URL:https://pretalx.devconf.info/devconf-us-2026/talk/7CPYUJ/
END:VEVENT
BEGIN:VEVENT
UID:pretalx-devconf-us-2026-3FHZHW@pretalx.devconf.info
DTSTART;TZID=EST:20260925T090000
DTEND;TZID=EST:20260925T102000
DESCRIPTION:Prompt optimization is still mostly hand-crafted: edit the word
 ing\, rerun a few examples\, inspect failures\, and repeat. That is like h
 and-crafting image kernels for every new vision problem. It can work for s
 imple cases\, but it does not scale well\, and it is hard to reproduce.\n\
 nGEPA offers an algorithmic alternative. It is a gradient-free form of RL 
 for LLM systems: sample candidate behaviors\, evaluate them\, use feedback
  from failures\, and search for better policies. But instead of updating m
 odel weights with PPO or GRPO\, GEPA optimizes the text artifacts that sha
 pe model behavior: prompts\, and system instructions.\n\nThis talk is for 
 engineers building LLM applications or agent workflows who want to move fr
 om prompt tweaking to evaluation-driven optimization. We’ll walk through
  the core GEPA recipe: writing an evaluation callable\, returning both rew
 ard and natural-language feedback\, exposing one or more text variables to
  optimize\, and running the optimizer on a small dataset. \n\nAttendees wi
 ll leave with a runnable notebook\, a mental model for when GEPA is useful
 \, and a reusable recipe for applying automatic prompt optimization to the
 ir own LLM systems.
DTSTAMP:20260727T165239Z
LOCATION:107 (Capacity 20)
SUMMARY:The Cheapest Way to RL tune LLMs: Getting Started with GEPA - Rohan
  Awhad
URL:https://pretalx.devconf.info/devconf-us-2026/talk/3FHZHW/
END:VEVENT
END:VCALENDAR