BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.coscup.org//coscup-2026//speaker//KWSYWJ
BEGIN:VTIMEZONE
TZID:CST
BEGIN:STANDARD
DTSTART:20000101T000000
RRULE:FREQ=YEARLY;BYMONTH=1
TZNAME:CST
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-coscup-2026-SVZZQG@pretalx.coscup.org
DTSTART;TZID=CST:20260808T103000
DTEND;TZID=CST:20260808T110000
DESCRIPTION:Large Language Models may start in frameworks like PyTorch\, bu
 t getting them to run efficiently on CPUs\, GPUs\, and specialized acceler
 ators increasingly depends on compiler infrastructure.\n\nAs machine learn
 ing systems have grown more complex and hardware has become more specializ
 ed\, the compiler layer has become a key bridge between model code and mac
 hine execution.\n\nThis talk introduces AI compilers from a practical pers
 pective\, covering why they matter\, what kinds of problems they solve\, a
 nd how open-source projects such as LLM MLIR and OpenXLA/XLA fit into the 
 broader stack. \nIt will also introduce a bit about tinygrad\, which is tr
 ying to take another approach rather than LLVM\, to compiler models runnin
 g on the hardware..\n\nIt also explains how ideas such as intermediate rep
 resentations\, lowering\, and backend targeting connect frameworks like Py
 Torch to real hardware.\n\nA small demo based on PyTorch and various compi
 lers (e.g. LLVM and tinygrad) helps make the overall pipeline more concret
 e and easier to understand.
DTSTAMP:20260713T132600Z
LOCATION:TR212
SUMMARY:From PyTorch to Hardware: An Introduction to LLM Compilers - Tommy 
 Han
URL:https://pretalx.coscup.org/coscup-2026/talk/SVZZQG/
END:VEVENT
BEGIN:VEVENT
UID:pretalx-coscup-2026-JLQPTY@pretalx.coscup.org
DTSTART;TZID=CST:20260809T094500
DTEND;TZID=CST:20260809T101500
DESCRIPTION:AI tools are making it easier than ever to generate code\, expl
 ore ideas\, and open pull requests quickly. But faster output does not alw
 ays lead to better PRs. In practice\, AI-generated changes often come with
  familiar problems: larger diffs\, weaker context\, less clear intent\, an
 d higher review overhead.\n\nIn this talk\, I will share a practical view 
 of where AI genuinely helps in day-to-day development\, where it starts to
  create problems\, and why I still choose to manually craft most pull requ
 ests. Rather than arguing against AI\, this session focuses on a more usef
 ul question for engineers and open-source contributors: how do we use AI a
 s a tool without giving up ownership\, clarity\, and reviewer empathy?\n\n
 I will also briefly use my own open-source contribution experience\, inclu
 ding work related to Zed\, as an example to show why a reviewable\, well-s
 tructured change is often more valuable than a quickly generated one. \n\n
 The goal of this talk is to provide practical suggestions about how to bui
 ld a proper PR to contribute to open-source in a healthy way\, and underst
 and why keep context understanding personally is important.
DTSTAMP:20260713T132600Z
LOCATION:RB105
SUMMARY:When AI Helps\, and Why I Still Write the PR Myself - Tommy Han
URL:https://pretalx.coscup.org/coscup-2026/talk/JLQPTY/
END:VEVENT
END:VCALENDAR