BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Palaner//palaner Single Event//EN
METHOD:PUBLISH
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:luma-bea1091584@palaner.app
DTSTAMP:20260727T060113Z
DTSTART:20260803T173000Z
DTEND:20260803T190000Z
SUMMARY:AI Research Night — Case Studies in Pretraining
LOCATION:Shoreditch Treehouse
DESCRIPTION:Saved in Palaner as: saved\n\nPalaner match: 17% · why: matches
  your usual pattern\n\nwith ElevenLabs\, Geodesic Research & PostHog · at 
 the Shoreditch Treehouse\nIf you've pretrained a model before you know you
  need data worth learning from\, compute\, and a serious answer to "why no
 t just call an API?" This is an evening with three teams who have answers 
 — at different scales — followed by an open discussion and plenty of canap
 és to keep us going.\nTalks (15 min each + Q&A)\nPretraining a GPT on a la
 ptop — Angelos Perivolaropoulos\, (Research Engineer at ElevenLabs) - What
  intuition from 10M parameters does (and doesn't) transfer upward.\nA from
 -scratch encoder for session replay data — Nico Waltz\, (AI Research Engin
 eer at PostHog) - A self-supervised encoder trained directly on DOM record
 ings\, built to replace a video-rasterizer + VLM pipeline in production.\n
 Alignment pretraining — Cameron Tice (Co-Founder & Executive Director\, Ge
 odesic Research) - Lessons from training 6.9B-parameter LLMs on 500B token
 s across four different data conditions to test how AI discourse in pretra
 ining corpora affects alignment.\nWho this is for​Active PhD students and 
 AI researchers with hands-on experience applying their work — people who h
 ave trained models. Registration is approved manually.\nAgenda\n18:30 — Do
 ors\, drinks\n19:00 — Talks\n20:10 — Open discussion\n20:40 — Dinner in th
 e treehouse
URL:https://luma.com/pretraining-night
CATEGORIES:palaner,saved
STATUS:TENTATIVE
END:VEVENT
END:VCALENDAR