Product Overview

Sengi

A search engine that filters noise and surfaces reliable information using AI-assisted ranking.


Problem

The problem with search today

Most search engines optimize for engagement, not accuracy. Users see results based on ad revenue, SEO manipulation, and popularity metrics — not relevance or trustworthiness.

This creates two issues:

For everyday queries, this is tolerable. For research, learning, or decision-making, it's a real problem.

Solution

What Sengi does differently

Sengi re-ranks search results using a quality-first approach. Instead of optimizing for clicks, we prioritize:

The goal is not to replace existing search infrastructure, but to add a filtering layer that improves signal-to-noise ratio for users who need reliable information.

Architecture

How it works

1. User submits a query

2. Query is processed and sent to multiple search APIs

3. Results are aggregated and deduplicated

4. AI model evaluates each result for quality signals

5. Re-ranked results are returned with source transparency

No proprietary crawling. No index ownership. Sengi is a processing layer, not a search engine replacement.

Google Cloud Usage

Infrastructure

Cloud Run — Stateless container deployment for API and frontend. Scales to zero when idle.
Vertex AI — Result quality scoring and summarization using Gemini models.
Cloud Storage — Static assets and anonymized query log archival.

Architecture chosen for cost efficiency at low scale and seamless scaling under growth. GCP is the natural fit for a containerized AI application.

Project Status Early Stage

Current state

This is an honest early-stage project. The product exists and runs, but is not production-ready for mass adoption.

Founder

About

Solo technical founder with a software engineering background. Building Sengi as an independent project to explore AI-assisted information retrieval.

This is a real project by a real person — not a pitch deck or a concept.