BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//programme.europython.eu//europython-2026//speaker//8TWA3
 E
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-europython-2026-CWBA3L@programme.europython.eu
DTSTART;TZID=CET:20260716T103000
DTEND;TZID=CET:20260716T111500
DESCRIPTION:OpenAI's embedding API is the default for most Python RAG appli
 cations\, but it creates real problems at scale: data privacy (every docum
 ent leaves your network)\, cost (per-token pricing compounds fast)\, no cu
 stomization (cannot fine-tune or swap models)\, and no quality validation.
  Most teams have never measured whether text-embedding-3-large actually wo
 rks well for their domain.\n\nOpen embedding models like BGE-M3\, GTE-Qwen
 2\, and E5 now match or beat OpenAI on MTEB benchmarks. The problem is not
  model quality anymore\, it is infrastructure: running these models with p
 roper batching\, memory management\, and quality assurance.\n\nThis talk w
 alks through a practical migration. We start from a typical LangChain app 
 calling OpenAI\, switch to self-hosted open models using SIE (an open-sour
 ce inference server with an OpenAI-compatible endpoint)\, validate that re
 trieval quality is maintained using MTEB\, and then go beyond what OpenAI 
 offers by adding sparse embeddings for keyword\, recall\, and cross-encode
 r reranking for precision.\n\nThe talk is aimed at Python developers who c
 urrently use OpenAI embeddings and models and want to understand their opt
 ions. No prior experience with model serving is required.
DTSTAMP:20260524T121930Z
LOCATION:Chamber Hall B (S3B)
SUMMARY:Self-Hosted Small Models: From OpenAI Lock-In to Open Models - Fili
 p Makraduli
URL:https://programme.europython.eu/europython-2026/talk/CWBA3L/
END:VEVENT
END:VCALENDAR
