Skip to content

Legible - I built an on-device document helper for immigrants using Gemma 4

6.4 relevance
Score Breakdown
technical depth
7
novelty
6
actionability
6
community
5
strategic
6
personal
8

Scored daily by a customisable AI persona to surface the most relevant engineering leadership news.

Practical AI app using Gemma 4, relevant.

2026-05-16 general Dev.to
Legible - I built an on-device document helper for immigrants using Gemma 4
Summary

Legible is an on-device document helper for immigrants that uses Gemma 4 E4B (via Ollama) to process photos of official documents locally—no cloud API, no data leaving the machine. The FastAPI backend streams structured XML output from the model, providing plain-language explanations, deadline countdowns, and next steps in 11 languages. Past scans are stored as Fernet-encrypted files, and the frontend is a single HTML file with no build step.

Key Takeaway

Evaluate Gemma 4 via Ollama for on-device document understanding; its structured output capability and offline operation make it ideal for privacy-critical multilingual apps.

Why it matters

This demonstrates a practical, privacy-first architecture for offline multimodal LLM applications using lightweight models, directly relevant to building secure, accessible AI tools for sensitive user data.