A line drawing of a row of electrical transmission pylons receding into the distance, drawn in amber light against a dark sky.

Jewish Education AI

The electric grid for
Jewish education

A shared, non-profit infrastructure that gives every Jewish learner — in any community, at any age — a study partner. Built once for the whole field: every institution shapes what it teaches, and the grid itself takes no side.

Oneshared technological core
Twoin class, and beyond it
Everysetting of Jewish life
6–90the ages it accompanies

01 The Challenge

Jewish learning is not the transfer of knowledge — it is the slow shaping of a person.

It happens through dialogue, relationship and challenge. Three structural constraints now stand between that and most Jewish learners — and the clock on AI is already running.

i

Cost

Day-school tuition has climbed beyond the reach of a widening share of families, forcing communities to choose between quality and access.

ii

Educators

There are not enough qualified Jewish educators, and the ones in place are stretched thin by the routine work of differentiating for every learner in the room.

iii

Reach

Millions of young Jews in the geographic periphery are simply unreached. No teacher-dependent model has ever been able to get to them.

AI is the first credible answer to all three at once

No previous technology has been able to address all three constraints at once. This one can.

  • On cost

    Personal tutoring at a cost per learner a community can actually carry, rather than a privilege that tracks tuition.

  • On educators

    Routine differentiation handled by the system, so scarce educators spend their hours on the work only a person can do.

  • On reach

    A learner in a community of forty Jews gets the same companion as one in a large day school.

But the AI that arrives by default is not built for this

Every one of those three answers assumes an AI built for Jewish learning. That is not the AI that is arriving. Off-the-shelf commercial AI is optimised for test drills and engagement metrics. It is very good at what it was built for, and it was not built for us: it has no idea that Jewish learning happens by arguing.

The core architecture and safety norms of educational AI are being set over the next two years, and independent development is beyond the reach of any single school. Left to the market, the wealthiest schools buy in and the rest do without — and the ones who do buy in will be teaching Judaism through a product built by people who never asked what it needs.

So the answer is not less AI. It is AI built from the ground up around Jewish pedagogy — by the field, for the field.

The Jewish world can build its own shared infrastructure today, or inherit someone else's tomorrow.

02 The Idea

Not another product. A utility.

No community builds its own power station. It plugs into a grid — shared, regulated, engineered once to a standard no single participant could afford alone. JEDAI applies that logic to Jewish learning.

A line diagram: one generating station with a transmission pylon on the left, sending amber lines out to a school, a camp cabin, a domed community building, a campus hall, and a single remote house on a hill.
The grid describes what JEDAI is: one core, generated once, drawn on by many institutions, shaped by each of them, owned by none. It says nothing about what travels through it — and that part is not a commodity. The grid is what gets built. Chavruta is what it carries.

Built once, for everyone

A single high-performance core — pedagogy, safety architecture, personalisation engine, avatar interface — engineered to a standard no individual institution could reach on its own.

Indifferent to subject

The core is domain-agnostic. It does not encode one curriculum; it carries whichever the field loads onto it. Hebrew and Holocaust education are being built first, and everything after them uses the same infrastructure.

Plug in permanently

Any school, camp, campus, youth movement or congregation connects to the same core and stays connected. The capability does not have to be rebuilt for each institution that joins.

03 How Learners Meet It

One avatar. In class, and beyond it.

JEDAI is a single avatar that grows with the learner. It begins inside the classroom, where the teacher stays in command — and keeps going, past the classroom, for the rest of a learning life.

A line illustration of a classroom: a teacher at the board, and rows of students each working with a small screen showing a different pattern, every screen wired into one shared line running along the floor.

Tier one · In classrooms January 2027

Classroom Tutor

A personal tutor for every student — and a personal assistant for every teacher.

No one can differentiate a single lesson thirty ways in real time. The Classroom Tutor is how one teacher reaches every learner in the room individually: each student working at exactly their own level, inside the lesson their teacher designed. It does not stand in for the teacher. It extends the teacher's reach to every student at once.

  • Scope. Built domain by domain, with the organisations that hold authority in each — Hebrew and Shoah education first.
  • Authority. The teacher sets the goals, the pace and the boundaries; the tutor works inside them.
  • Reach. Wherever educators are present, the system extends their reach rather than replacing it — absorbing routine differentiation so their hours go to mentorship.
  • Status. Running on a proven engine; the first Jewish content domains enter classrooms in January 2027, across North America, Latin America, Europe and South Africa.
A line illustration of two learners facing each other across a table with an open book between them, mid-argument, while further pairs in the background are linked to them by arcing lines.

Tier two · In development

Chavruta

A study partner for life — one that argues back.

Chavruta is the heart of JEDAI. It teaches through dialogue rather than drills: questioning, pushing, conceding, and deliberately taking the opposing side — the way a real study partner does. It does not deliver content to a learner. It thinks with them.

  • Lifelong. One companion from age six to ninety, carrying what it has learned about a learner forward across years.
  • Everywhere. The same partner in the classroom, at camp, on a trip to Israel, on campus, and long after.
  • Connective. It also matches isolated learners anywhere in the world into chavruta with one another — the pairing that geography has always denied them.

The moments nobody is scheduled for

Much of what forms a Jewish life happens off the schedule: the night before a first trip to Israel, the argument in a dormitory corridor, the question on the walk home from camp. Even the finest educator — and the field has many — cannot be there at two in the morning, or in thirty places at once.

A nineteen-year-old, a week before her first trip to Israel, is watching the news and quietly deciding she might not go. Her hesitation has nothing to do with logistics. It is about walking into a conflict she does not feel equipped to understand, and whether going means taking a side she has not thought through. Her programme's materials are thorough, but the next session is days away and this feeling is not waiting. So she does what anyone her age would do at midnight: she opens a general-purpose chatbot and asks it to explain the conflict. It gives her a fast, confident answer flattened into a story of good and bad. She now has an answer. She does not have the understanding she needed, and the doubt that almost made her cancel has not moved.

A Chavruta would meet her in exactly that moment — as part of her own programme, drawing only on material her own educators had vetted. It would register the anxiety and the real question underneath it, slow down, and hold the competing narratives open rather than pick one. And it would not forget. The same conversation would carry through the trip, and long after she is home. Where a chatbot closes a query, a chavruta opens a relationship.
  • Journeys to Israel

    Preparation in the weeks before departure, accompaniment while they are there, and reflection once they are home — so a formative ten days stops evaporating on the flight back.

  • Camp, movement and community

    The intensity of a summer has always been followed by eleven quiet months. A companion extends the relational work of counsellors, hosts and rabbis into the gaps between gatherings.

  • Campus and early adulthood

    The years in which Jewish questions become most pressing are the years in which institutional Jewish education mostly stops. A partner who already knows a learner does not start over at eighteen.

  • Everyone outside the walls

    Unaffiliated families, interfaith households, isolated communities, adults who were never given a Jewish education and will not enrol in one now. Almost nothing reaches them today. A companion can.

Two things have always limited Jewish education: how many gifted educators exist, and where a learner happens to live. A companion bound to no room, no timetable and no country does not ease those limits — it removes them.

Dialogue, not drills

Jewish learning is a relationship, not a transaction. Everything in the system is built to that.

  1. 01

    It asks before it answers

    The default move is a question that opens the problem, not a fact that closes it.

  2. 02

    It takes the other side

    Machloket — structured disagreement — is a method, not a malfunction. The partner is built to press.

  3. 03

    It remembers

    Continuity is the point. A partner who recalls what you struggled with three years ago can teach you things a fresh tutor never could.

  4. 04

    It works inside declared values

    Guardrails are set by educators and made explicit, not inferred from an engagement objective.

  5. 05

    It defers to human beings

    Where a teacher, parent or rabbi belongs in the loop, the system routes toward them rather than around them.

04 Technology

Adaptation, not speculation.

JEDAI is not an experiment in whether personalised AI tutoring can work. That question has already been answered, at national scale, in a different subject. JEDAI takes a computational core with a proven record and rebuilds the pedagogy on top of it.

A foundation, not a prototype

JEDAI does not begin from an empty screen. Its computational core is the Multi-Parameter Customization engine — MPC — developed by I-Teach, an Israeli learning-technology company. The engine has been in continuous operation since 2018 inside Israel's national education system.

Scale was not what decided it. Two requirements did: a study partner built from the ground up around Jewish pedagogy rather than a content-delivery product adapted to it, and a permanent commitment to at-cost pricing. I-Teach met both, and agreed to build the result as shared infrastructure owned by the field — not a platform owned by a vendor.

The method also predates the software. Twenty years of classroom teaching came first; the engine came after. What runs underneath JEDAI is a pedagogy tested on children long before it was written in code — not a course wrapped around a language model.

2018 in continuous national service since
700+ partner schools in Israel
96.4 average matriculation score reached by students who had entered well behind grade level — against an 84.6 national average, in roughly 60% of the usual time. Independently verified at the Technion.
98% learner retention, with a 4.91-out-of-5 satisfaction score across more than 83,000 learners who completed full programmes

These figures describe the MPC engine's record in Israel's national school system — the proven foundation JEDAI builds on. They are not JEDAI outcomes. JEDAI's own results begin with the first pilot in January 2027.

How the engine works

Most educational software adapts to answers: get it wrong, get an easier question. MPC adapts to the learner. It maintains a continuously updated profile of each individual and adjusts in real time — not once a term, but inside the conversation as it happens.

  • 01

    It tracks about twenty dimensions of a learner

    Not just right and wrong: knowledge gaps, pace, persistence, interests, goals, learning style — seventeen of them in the Classroom Tutor, about twenty in Chavruta. A capable learner who is drifting needs the opposite of a confident one missing a foundation, and the engine can tell them apart.

  • 02

    It adjusts three things in response

    Not only what comes next, but the tone it comes in and the kind of feedback that follows. The same idea reaches a hesitant learner and an impatient one in two different registers.

  • 03

    The profile persists

    It accumulates across years and settings rather than resetting each session.

  • 04

    It is indifferent to subject matter

    The engine was proven on mathematics and science. Nothing in it is mathematical. It can carry Hebrew, sacred text, Israel and peoplehood — and whatever the field adds after them.

What JEDAI builds is the part that does not exist yet: the Jewish pedagogy, the values architecture, the guardrails, the domain content, the chavruta interaction model, and the governance that keeps all of it in the field's hands.

05 Governance

Trust is an architecture, not a promise.

Field-wide adoption depends on something stronger than good intentions. JEDAI is built so that it structurally cannot behave badly toward the institutions that rely on it.

An independent public-benefit company

JEDAI is being established as an independent non-profit public-benefit company (חל״צ) — a legal form whose purpose is the public good rather than returns to owners.

No one can be locked out

Institutions with deep expertise lead the fields they know best, and that leadership is real. What it never includes is a veto: no partner can shut another institution out of a field of learning, restrict what it teaches, or impose its reading on anyone else's learners.

An educator-led Vision and Design Council

Pedagogy, content, values boundaries and the research agenda are set by the Educational Vision and Design Council: practising Jewish educators and scholars from across the movements. Its published principles apply to every domain on the grid.

It cannot become your rival

JEDAI is not a commercial venture and has no market to win. It does not recruit learners, sell to your constituents, or profit from your dependence on it.

What belongs to whom

The pedagogy is the field's: curricula, learning objectives, values architecture, guardrails. Learning records stay open and portable — including to another provider. At-cost pricing holds beyond the life of any single engagement. The engine and the code belong to those who build them; JEDAI holds no technology and wants none.

Judged from the outside

Outcomes are measured by an independent evaluator appointed by JEDAI — proficiency, and whether formative learning is actually taking place.

Behaviour, never biometrics

The engine records what a learner did with a problem — the time taken, the attempts made, the kind of error, which explanation finally worked. Persistence and engagement are read from those actions alone. Nothing from face, voice or physiology, which the EU AI Act prohibits in education.

Who sets the boundaries when no institution does

Through a school, a camp or a movement, that institution's own educators decide. For a learner who comes alone — an unaffiliated adult, an interfaith family, a teenager in a community of forty Jews — three layers:

  • A published baseline. Set by the Educational Vision and Design Council. Pluralistic, age-appropriate, built to hold competing Jewish readings open rather than adjudicate between them.
  • Or a community's profile. A learner may opt into one deliberately. Never assigned by default.
  • A safety floor, for everyone. Age-appropriate content. Nothing read from face or voice. A route to a human being when a conversation turns from learning to distress.

No margin, ever

Institutions pay only the real, ongoing cost of running the grid — infrastructure, computing, maintenance and continued development. Never a markup, never a profit margin, and never licence fees that grow as the field's dependence on it grows.

A dark map of many small points of amber light in loose clusters, joined by long arcs that reach out to isolated single points at the edges.

06 Partnership & Reach

A grid is only worth what it reaches.

JEDAI reaches learners through the organisations that already hold their trust. Your curriculum, your name, your learners — carried further than a programme calendar allows.

An umbrella organisation runs a superb ten-day programme for thousands of young adults a year. Two decades of curriculum work, and it shows: people arrive uncertain and leave changed. Then, every year, the same thing. They turn up barely prepared, and within a fortnight of landing home everyone is gone — not for want of content or conviction, but for want of anything that scales to thousands of individual people.

That gap is what JEDAI closes: preparation before, accompaniment during, continuity after, from a companion that already knows them. The organisation hands its relationship to no one. It extends it.

Your content

Your educators decide what is taught, which texts, which framings. JEDAI's pedagogy is dialogic; the positions in the room are yours.

Your name

The avatar sits inside your own programme, under your own branding. Participants meet it as yours, because it is.

Your domain

Every field of learning is built with the organisations that hold authority in it. Partners design their domain; they do not receive it.

Not a rival

JEDAI has no learners of its own to win and nothing to sell your constituents.

Built for organisations like these

  • Day schools
  • Supplementary schools
  • Congregations
  • Campus organisations
  • Summer camps
  • Youth movements
  • Israel-experience providers
  • Federations and community centres
  • Museums and memory institutions
  • Diaspora communities worldwide

Built to carry fields of learning like these

  • Hebrew
  • Shoah education and memory
  • Sacred and rabbinic text
  • Israel education
  • Jewish peoplehood
  • Heritage languages and traditions
  • Adult and family learning
  • Educator development
  • Digital literacy and antisemitism
  • and whatever the field brings next

Hebrew and Shoah education are in build now, launching January 2027. Everything after them runs on the same infrastructure.

  1. Now

    Classroom Tutor in build

    The engine is proven and running. Hebrew and Shoah education enter classrooms in January 2027, across four continents.

  2. In build

    Chavruta

    Roughly eighteen to twenty months of development to bring the lifelong dialogic companion to launch.

  3. Then

    Field-wide scale-up

    Progressive rollout across the partner network and the full set of content domains.

  4. Ongoing

    Renewal of the core

    Successive platform generations, so the grid keeps pace with the technology beneath it.

07 Stewardship

Who stands behind it.

Two teams, two distinct responsibilities. JEDAI sets the educational priorities, the values alignment and the work with institutions in the field. I-Teach builds and operates the computational core. The separation is deliberate: neither side can quietly become the other.

JEDAI

Educational vision, values architecture, governance, and implementation with institutions across the field.

  • Dr. Shivi Greenfield
    Founding Director
  • Tova Peretz
    Director of Operations
  • Rabbi Craig Kacev
    Director of Education
  • Rabbi Tully Harcsztark
    Co-Chair, Educational Vision and Design Council

I-Teach

Engineering leadership behind the Multi-Parameter Customization engine. Unit 8200 and Technion alumni.

  • Yair Cohen
    Executive Chairman
    Former commander of Unit 8200. Decades building and scaling deep-technology ventures.
  • Yosi Levi
    President
    Built Teachin over twenty years. Inventor of the MPC method.
  • Boaz Bryger
    Chief Executive Officer
    Mathematics teacher turned technology executive. Previously Qualcomm.
  • Jenny Gurevich
    Pedagogical Director
    Mathematics educator and teacher trainer. Leads classroom pedagogy.
  • Dr. Ron Davidson
    Chief Technology Officer
    PhD, Stanford. Architect of scalable AI systems. Previously Skybox (Google).
  • Alon Fliess
    VP Research & Development
    Microsoft MVP. Thirty years in software and applied AI.
  • Eva Shelley
    Product Manager, PMO
    Translates deep technology into classroom value.

Division of roles: I-Teach builds and runs the engine. JEDAI decides what the engine is for — the pedagogy, the values boundaries, the content priorities, and the terms on which institutions plug in.

Domain leadership: Every content domain has its own pedagogical lead — the authority who sets the learning objectives, the assessment logic and the rules the engine works under, and who approves quality before anything reaches a classroom.

08 Get Involved

Built once. Then it belongs to the field.

The architecture of educational AI is being settled now, by people who are not asking what Jewish learning needs. There is a short window in which the Jewish world can build its own, on its own terms. It will not stay open.

JEDAI is being built with the schools, umbrella organisations, educators and researchers who want a say in what a shared Jewish learning infrastructure becomes. If your organisation has learners to reach, content worth carrying, or a view on how this ought to be done — that conversation is open.

JEDAI (in formation) · An initiative of the Pincus Fund for Jewish Education · Jerusalem