BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//pyconde-pydata-2026//speaker//UXME8R
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-pyconde-pydata-2026-GAUNKM@pretalx.com
DTSTART;TZID=CET:20260416T113500
DTEND;TZID=CET:20260416T122000
DESCRIPTION:E-commerce cataloging at idealo operates at extreme scale: 4.5 
 billion offers from 50\,000+ shops across six countries\, with peak ingest
 ion rates of 4.8 million offers per minute. While large language models (L
 LMs) provide strong classification accuracy\, they are too slow and costly
  for billion-scale real-time processing. This talk shows how idealo builds
  a cost-efficient\, high-throughput machine learning system that leverages
  LLM knowledge without deploying full models in production. \n\nWe present
  how knowledge distillation from a large e5 instruction model enables a co
 mpact multilingual MiniLM encoder to achieve high accuracy\, and how optim
 ized inference runtimes and specialized hardware such as AWS Neuron help m
 eet strict latency and cost requirements. Beyond modeling\, we highlight k
 ey operational challenges: constructing training datasets from massively i
 mbalanced data\, selecting the right encoder architecture from today’s m
 odel landscape\, and designing a robust MLOps lifecycle with automated dat
 a sampling\, training\, deployment\, and monitoring. \n\nAttendees will le
 arn practical techniques for scaling ML systems under real-world constrain
 ts\, how to extract value from LLMs when they are too large to serve direc
 tly\, and how to transition research prototypes into reliable\, high-volum
 e production pipelines.
DTSTAMP:20260412T141732Z
LOCATION:Palladium [2nd Floor]
SUMMARY:When LLMs Are Too Big: Building Cost-Efficient High-Throughput ML S
 ystems for E-Commerce Cataloging - Tobias Senst\, Bastian Wandt
URL:https://pretalx.com/pyconde-pydata-2026/talk/GAUNKM/
END:VEVENT
END:VCALENDAR