About — Maaty Frankelson & MF Smart Research

A historian who builds the technology he uses

MF Smart Research was founded by Maaty Frankelson — a researcher, archivist and Ph.D. candidate in History — to bring modern artificial intelligence into the daily practice of historical research.

The combination is unusual on purpose: most AI consultancies do not understand archival sources, and most historians do not build their own tools. MF Smart Research operates at the intersection. Every model is shaped by primary-source experience, and every interpretation is grounded in a researcher's instincts about provenance, context and reliability.

What we do

We design and operate AI systems that handle four problems that historians have lived with for centuries:

  1. Reading what no one can read. Custom OCR and Handwritten Text Recognition (HTR) for Hebrew, Yiddish, Ladino, Aramaic, Arabic, Ottoman Turkish, Polish, German (including Kurrent and Sütterlin), Russian and English — across rabbinical hands, community ledgers, personal diaries, court records, ship manifests and degraded newspapers.
  2. Finding what's lost in volume. Retrieval-Augmented Generation (RAG) and autonomous research agents that search across millions of pages of multilingual sources to surface the one document a researcher actually needs.
  3. Connecting what's scattered. Knowledge graphs that link people, places, events and dates across collections that were never designed to talk to each other.
  4. Telling the story. Annotated transcriptions, scholarly footnotes, Visual data mining of photo and film archives, and interview-based documentary research.

Audiences we serve

  • Universities and academic researchers — AI-assisted qualitative research at scale, methodology consulting and integration with institutional databases.
  • Archives, libraries and museums — end-to-end digitization with custom HTR and metadata enrichment for collections of varying scale.
  • Documentary filmmakers and investigative journalists — visual archive mining, interview transcription, fact verification.
  • Independent scholars and private clients — handwriting decoding, family archive reconstruction, Holocaust documentation, immigration tracing.

Selected work

Completed and ongoing engagements span several languages and source types, including:

  • In-depth academic research with university researchers in sociology and adjacent humanities.
  • OCR and HTR pipelines for early-20th-century Yiddish printed and handwritten materials — periodicals, correspondence and personal documents.
  • International projects documenting the history of Jewish communities across multiple countries, combining cross-archive source work with structured knowledge bases.
  • Semantic-search environments and annotated transcriptions delivered for individual scholars and small institutions.

Methodology

Every project goes through four stages:

  1. Scoping. Free 30-minute consultation. We assess sources, expected accuracy, timeline and the realistic value AI can add (sometimes the answer is "less than you think" — and that matters).
  2. Pilot. A small, paid pilot — usually a few hundred pages — to calibrate models and validate accuracy on the actual material. Results are shared transparently with error rates and confidence scores.
  3. Production. Full-scale processing with quality assurance loops, human-in-the-loop review on uncertain passages and source attribution on every output.
  4. Handover. Searchable archive, structured metadata, source-cited reports, training for the institution's staff, and optional ongoing support.

We never publish or use client material without written consent. Sensitive collections (Holocaust testimonies, family records, legal documents) are processed under formal agreements that meet institutional and ethical standards.

Languages and scripts

Hebrew (biblical, rabbinic, modern, handwritten and printed), Yiddish (multiple regional traditions), Ladino / Judeo-Spanish (Hebrew and Latin script), Aramaic, Arabic, Ottoman Turkish, German (including Kurrent and Sütterlin), Polish, Russian (pre- and post-reform Cyrillic), French and English.

Founder

Maaty Frankelson (מתי פרנקלזון)

Researcher, archivist, Ph.D. candidate in History. Combines training in archival science with hands-on engineering of AI pipelines for historical sources. Available for talks, sponsored research, institutional consulting and joint grant applications.

Press, talks and publications

This section will be populated as engagements take place. To invite Maaty for a talk, panel, podcast, peer review or guest article, write to [email protected].

Get in touch

We answer every serious inquiry, whether you are a national archive scoping a five-year program or a grandchild trying to read one box of letters in a drawer.

Request a free 30-minute consultation →

Start Your Research Project