AI Lab
For Jewish Memory
Using AI to survey, read, and catalog the world’s Jewish records—before they’re lost—so every family can access their heritage and build resilience against rising antisemitism
Problem → Solution → Outcome
- Problem: 85% of Jewish records still exist yet are vulnerable; manual survey averages $84 and 4.7 hours per JCat entry—unsustainable at global scale.
- Solution: Train multi‑language HTR models and automate discovery, assessment, and JCat cataloging.
- Outcome: Faster coverage, town‑based search in JCat, and accessible heritage experiences for researchers and the curious.
How the AI Lab Works
4 Pillars
Discovery Automation
triage large digitized sets, summarize contents
Handwriting Recognition
models in Russian, Polish, Yiddish, Hebrew, Arabic; Ladino advancing; more to follow
Auto‑Catalog for JCat
structured descriptions & metadata searchable by town
Experiences & Apps
“PastPort” & “Ask Rosenberg” Jewish Roots Journey make discovery intuitive
Impact & Momentum
- 5+ years building multi‑language HTR capacity.
- Major LDVDF ground‑truth contribution to Transkribus’s new Polish SuperModel (handwritten + printed) — “Polski Bizon”.
- Pioneering public Yiddish HTR models launched at IAJGS 2025 (Ft. Wayne). Learn More…
- Collaborations: Pinkas of Brody (public model), Baghdad 1951 migration list (40k names), Ukrainian local‑history project (Purdue visiting scholar).
See It in Action
JCat — Search by town, discover what exists.
Reading Bubbe’s Letters with AI: Unlocking Yiddish & Jewish History
PastPort (Preview) — 90‑sec walkthrough of name‑to‑town journey
Support the AI Lab
❤️ Help fund this work
L’Dor V’Dor Foundation is a 501(c)(3) nonprofit. Your donation helps us:
- Discover vulnerable Jewish record collections before they’re lost
- Create the expert “ground truth” that AI needs to learn to read new scripts and languages
- Train and improve AI models in historic handwritten Yiddish, Hebrew, Russian, Polish, Arabic, Ladino, and more languages
- Integrate these tools into DoJR and JCat so families, scholars, and communities can actually FIND and USE this history
- Run workshops and share our models freely, multiplying impact far beyond our own team
👉 Donate to support this work: