Skip to content

README.md

Latest commit

 

History

History
24 lines (19 loc) · 1.33 KB

README.md

File metadata and controls

24 lines (19 loc) · 1.33 KB
Raw

Logo

sensa

Fighting cervical cancer misinformation in Romania with Gemma 4

The problem

Romania has the highest cervical cancer mortality rate in the EU, yet HPV vaccination uptake remains critically low — driven largely by misinformation spreading unchecked through social media in rural communities. There is no trusted, judgment-free, Romanian-language digital tool where a teenage girl can get accurate answers to the questions she's too embarrassed to ask anyone in her life.

What Sensa does

Sensa is a conversational health education assistant powered by Gemma 4. It speaks simple Romanian, debunks HPV and cervical cancer misinformation with cited medical sources, and is architecturally designed to run offline on a mobile phone using the E4B model, so a girl like Maria can access it without internet, without a server, and without anyone seeing her search history.

Architecture

Architecture diagram Four-layer pipeline: deterministic safety → myth RAG → constrained Gemma 4 E4B → response validation.

How to run

  1. Open sensa.ipynb in Kaggle
  2. Attach Gemma 4 E4B model from Kaggle Models
  3. Enable GPU T4 accelerator
  4. Run All

Attribution

Sensa is built on Gemma 4 E4B, developed by Google DeepMind. Sensa is not endorsed by or affiliated with Google. Gemma is a trademark of Google LLC.