BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//pydata-amsterdam2026//talk//EUY789
BEGIN:VTIMEZONE
TZID:CET
BEGIN:STANDARD
DTSTART:20001029T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-pydata-amsterdam2026-EUY789@pretalx.com
DTSTART;TZID=CET:20260911T115000
DTEND;TZID=CET:20260911T122000
DESCRIPTION:Pretraining foundation models on tabular and sequential data pr
 esents challenges that differ fundamentally from NLP or vision. This talk 
 covers the key design decisions involved in building a payments foundation
  model trained on billions of transactions and trillions tokens: tokenisat
 ion of heterogeneous features\, sequence construction\, masking strategies
  and pretraining objectives for sequential tabular data\, and the architec
 tural trade-offs between hierarchical and language-model framings. Attende
 es will leave with transferable techniques for adapting self-supervised pr
 etraining to non-text sequential data.
DTSTAMP:20260710T141215Z
LOCATION:Main stage
SUMMARY:Trillion-Token Pretraining: Building a Foundational Model for payme
 nt data - Hanna van der Vlis\, Martin Iglesias Goyanes
URL:https://pretalx.com/pydata-amsterdam2026/talk/EUY789/
END:VEVENT
END:VCALENDAR