Product Overview · Avodah Connect

AI language translation.

Role: Senior Product Designer
Team: Product Manager · Machine Learning Engineer
Tools: Figma

Problem

Some languages are at risk of going extinct — and with them, the cultural knowledge they carry. Traditional Bible translation into a single language takes 20–25 years, leaving most of the world's low-resource languages behind.

Making language available is important — yet decisions about meaning, theology, and use must remain firmly in human hands.

The Product

An AI-assisted Bible translation platform.

Avodah Connect is built for low-resource languages. Small language models, trained on each language's own materials, help translators produce a first-pass translation while detecting meaning, intent, and tone verse by verse. Translators, consultants, and elders then review, refine, and approve every passage — the AI assists, but humans always decide. The goal: compress translation work that has historically taken 20–25 years into four to five, without sacrificing accuracy or community ownership.

My Approach

A full re-architecture, validated week over week.

This was a full product re-architecture and a complete redo of the product and user experience. With competitive analysis already handled by a third party, the product manager and I moved straight from requirements into design, validating throughout in bi-weekly sessions with internal account managers who worked side-by-side with external customers every day.

A standard design system had been created just before I joined the project, so I added details, usages, and behaviors to existing components and created new ones as patterns surfaced — handing off detailed interaction specs that developers then built and implemented into the UI.

Solutions

Key views across the translation workflow.

The workflow below begins on the admin side — setting up a project, adding contributors, and tracking progress from the project dashboard. From there it moves to the contributor experience: opening the level dashboard, translating verse by verse, and managing progress as work moves from Ready to In progress to Complete.