Stratos

STARTOS is an AI-powered civic misinformation verification platform that combines Elasticsearch retrieval with AWS Bedrock-powered reasoning workflows.

Here’s how it works.

We first ingest and index trusted civic advisories from sources like BMRCL, BBMP, BWSSB, BESCOM, and Bengaluru Traffic Police into Elasticsearch.

When a user submits a viral claim, STARTOS performs semantic retrieval using Elasticsearch to fetch the most relevant institutional notices and operational advisories.

Those retrieved evidence documents are then passed into our AWS Bedrock verification workflow, where the system evaluates whether the claim is verified, misleading, false, or lacks sufficient evidence.

Let me demonstrate.

Here I paste: “Purple Line completely shut tomorrow across Bengaluru.”

Now STARTOS retrieves relevant metro advisories from Elasticsearch and compares the viral claim against institutional evidence.

As we can see, the system identifies that the claim exaggerates the actual advisory. The official notice only mentions scheduled maintenance and interval adjustments — not a complete shutdown.

The platform also surfaces the exact evidence sources used for verification, making the reasoning transparent and traceable.

Technically, the system combines:

Elasticsearch for retrieval and evidence ranking, AWS Bedrock for verification workflows, and Streamlit for rapid civic intelligence visualization.

Our vision is to evolve STARTOS into a real-time civic misinformation infrastructure layer for Bengaluru through live institutional integrations and continuously indexed public advisories.

Thank you.

Built With

• amazon-web-services
• elasticsearch
• streamlit