Mordechai Potash

AI Engineer × Torah Technology

Senior AI Engineer Application

I've already built on your API

14+ months of production work with Sefaria data. 3.9 million Hebrew sources. Real applications, real users.

3.9M
Hebrew Sources
3.8M
Citation Links
117K
Vector Embeddings
171
PDF Outputs

Sefaria Projects

Torah DuckDB

Complete

Production data pipeline ingesting the entire Sefaria corpus into a queryable analytical database with full-text search.

  • 5.4 GB database
  • 14.9M records
  • 3.8M citation links
  • 3.3M search index entries
DuckDBPythonSefaria API

Talmudic Study App

Deployed

Real-time streaming translation using Gemini generateContentStream(). Never trust Sefaria English—translate fresh from Hebrew.

  • Next.js 14 + TypeScript
  • Supabase backend
  • LRU caching (500 texts)
  • Zustand + React Query
Next.jsTypeScriptSupabaseGemini APISefaria API

Prism for Torah

Complete

Publishing pipeline transforming Sefaria texts into beautifully typeset bilingual PDFs. Built 3 months before OpenAI released their Prism.

  • 150+ chapters processed
  • 171 PDF outputs
  • XeLaTeX + SBL Hebrew
  • Tur, Nefesh HaChaim, Derech Hashem
PythonGeminiXeLaTeXSefaria API

MCP Sefaria Server

Complete

Model Context Protocol server providing LLM agents access to the Sefaria library. Tools for text retrieval, commentaries, search.

  • get_text, get_commentaries
  • search_texts, get_daily_learnings
  • Claude Desktop integration
PythonMCP SDKSefaria API

Ohr Avraham Chaim

Active

Multi-volume publishing project producing sefarim from Sefaria sources with AI-assisted translation.

  • 7 sefarim in progress
  • Full Hebrew + translation
  • Print-ready output
Sefaria APIGeminiLaTeX

AI Technologies Experience

Gemini Streaming

generateContentStream() for real-time translation

Vector Search

nomic-embed-text-v1.5, 256ms semantic queries

RAG Pipelines

353K messages embedded, searchable by concept

LangChain

Chaining LLM calls, retrieval workflows

MCP Servers

92 tools exposing Brain system to agents

Elasticsearch

Shiurim transcript search, Torah text indexing

React / Next.js

Multiple production applications

  • WOTCFY — Full tax automation platform
  • Talmudic Study App — Deployed production
  • Sparkii Command Center — Dashboard UI
  • React 18, App Router, shadcn/ui
  • Zustand, React Query, TailwindCSS

Supabase

5+ production projects

  • Auth (magic links, service role)
  • Storage (PDF uploads, buckets)
  • Edge Functions + secrets
  • PostgreSQL schemas
  • Realtime sync (Gmail → DB)
  • 5+ production projects

Why Sefaria

"I AM your user. I've built the tools I wish existed."

Sefaria's mission of making Jewish texts universally accessible isn't abstract to me — it's why I've spent 14 months building tools on your API. The Torah DuckDB, the publishing pipelines, the translation tools — all exist because I needed them for my own learning.

Full Tech Stack

Python
Next.js / React
TypeScript
Supabase
LanceDB
Elasticsearch
PostgreSQL / DuckDB
Claude / OpenAI / Gemini
LangChain
MCP Protocol
Docker
Kubernetes

✓ = Production experience • Hover for details

Production RAG System: Brain MCP

I've already built a production semantic search system for my own data — Brain MCP: ready to apply this infrastructure to Sefaria's corpus.

374K
Messages Indexed
117K
Vector Embeddings
256ms
Query Latency
92
MCP Tools
nomic-embed-text-v1.5LanceDBDuckDB AnalyticsSemantic SearchThinking Trajectory

Let's Talk

Happy to demo any of these projects live or discuss how I can contribute to Sefaria's mission.

mordechaipotash@gmail.comGitHub