Eyal Pasha
Get in touch Email

LegalTech AI

Hebrew NLP and LegalTech AI designed for production teams in Israel, Europe, and cross-border legal workflows.

A closer look at the LegalTech platform work: multi-model orchestration, document intelligence, and full-stack delivery for demanding Hebrew and English legal workflows.

By Eyal Pasha (אייל פשה) · Last updated April 9, 2026

The current production work is centered on an enterprise LegalTech platform where AI has to be useful under operational pressure, not just interesting in isolation, for teams handling local and cross-border matters.

That means building around actual legal workflows: high-volume document intake, clause-level risk review, structured extraction, predictive case context, and communication layers that fit how legal teams already work. The architecture spans OpenAI, Gemini, and local LLMs via vLLM, backed by Python services, production NLP pipelines, and a scalable Azure-hosted PostgreSQL foundation built for demanding enterprise workloads.

The Hebrew angle matters because off-the-shelf international tooling is rarely enough on its own. The value comes from aligning language nuance, business process, and deployment reality until the system is actually adoptable by teams working across Hebrew and English legal material.

In practice

  • Brings document intelligence, AI architecture, and delivery into one system.
  • Relevant for Europe and international teams dealing with Hebrew and English documents, Israel-linked operations, or multilingual legal workflows.
  • Shows how language nuance, workflow design, and infrastructure decisions fit together.
  • Reflects production constraints rather than demo-stage assumptions.

System shape

  • Multi-model orchestration across OpenAI, Gemini, and local LLMs via vLLM.
  • Python and TensorFlow pipelines for classification, extraction, and predictive legal analytics.
  • Azure PostgreSQL foundation supporting enterprise-scale legal document operations.
  • Microsoft 365, Microsoft Graph, and Microsoft Teams integrations for in-channel adoption.

Commercial logic

  • Reduce manual review load for legal teams.
  • Surface risk and structured information earlier in the workflow.
  • Keep AI inside existing operational channels rather than adding another disconnected tool.
  • Design systems that can survive real users, real budgets, and enterprise reliability expectations.

Core areas

Four parts of the work that define the system.

These are the four parts that shape the work in production.

NLP

Document intelligence

Information extraction, classification, and clause analysis built around enterprise legal documents rather than generic text tasks.

Architecture

Model orchestration

A multi-model AI stack where hosted and local models are selected according to task, cost, latency, and operational fit.

Infrastructure

Production persistence

Azure PostgreSQL, backend services, and full-stack product delivery are treated as part of the AI system, not as separate afterthoughts.

Adoption

Workflow fit

Integrations with Microsoft 365 and Microsoft Teams keep the product close to the way teams actually communicate, review work, and collaborate.

More to read

More on the broader AI work, research, and academic background.

A few deeper reads across the rest of the work.

Israel

AI/ML engineer in Israel focused on Hebrew NLP and applied ML systems.

A closer look at Israel-based AI/ML work across Hebrew AI, applied machine learning, and production systems.

Read more

Research

WMSM research on Hebrew Sign Language recognition.

A closer look at the WMSM publication, the key results, and the international research-to-product path into Handibur.

Read more

Academic

Academic AI research in Israel, from IEEE publication to interdisciplinary thesis.

A closer look at the publication record, international conference footprint, COLMAN background, and the interdisciplinary thesis.

Read more

Questions

A few things people usually ask about the platform.

The short version of the system, the language challenges, and what makes it hold up in production.

What kind of LegalTech AI work does this cover?

Eyal Pasha (אייל פשה) leads development of an enterprise LegalTech SaaS platform at Sensemaking Israel in Rehovot — the first AI-native operating system built specifically for Israeli law firms. The system handles clause risk review using custom AI agents, structured extraction from Hebrew legal documents, document intelligence across complex legal and medical files, predictive case analytics, and workflow integrations that keep AI embedded inside real legal operations rather than running as a disconnected tool.

Why does Hebrew NLP matter in this context?

Hebrew-language legal work requires specialized NLP beyond generic off-the-shelf tools. Hebrew has right-to-left directionality, complex morphological structure with root-based word formation, and significantly less training data available compared to English. Israeli legal documents mix Hebrew and English terminology, use domain-specific phrasing, and follow unique formatting conventions. The platform adapts language handling, tokenization strategies, model selection, and workflow logic specifically for these realities, ensuring accurate processing of Hebrew contracts, court filings, and regulatory documents used by Israeli law firms.

What makes the system production-ready rather than a demo?

The platform runs a multi-model AI orchestration stack combining OpenAI, Gemini, and local LLMs via vLLM, backed by Python NLP services, a scalable Azure-hosted PostgreSQL architecture designed for high-volume enterprise data, and deep integrations with Microsoft 365, Microsoft Graph, and Microsoft Teams — developed in close collaboration with Microsoft product teams. This is not a prototype; it handles real enterprise workloads for Israeli legal professionals processing Hebrew and English documents at scale daily.

Contact

Open to teams building AI for legal, enterprise, or Hebrew-language workflows.

If the work needs Hebrew NLP depth and enterprise delivery experience, this is a good place to start.

Email or LinkedIn is the fastest route.

Open to conversations about LegalTech AI, Hebrew NLP, enterprise document intelligence, and production AI systems for Israeli legal workflows.

eyal@eyalpasha.com LinkedIn GitHub IEEE ResearchGate Scholar ORCID