AI Systems / Active Build

Aletheia

Hybrid search platform running keyword and vector retrieval with reranking, evaluation pipelines, and ranking observability.

Active buildFlagshipPlatformPublished record

Project Dossier

Active BuildFlagship
Category
AI Systems
Type
Platform
Status source
Project model

A structured read on the build.

The page uses approved project data and keeps current-state boundaries visible.

Active Build

Aletheia runs keyword and vector retrieval side by side, reranks the merged results, and measures quality with evaluation pipelines and observability tooling. FastAPI services over PostgreSQL, OpenSearch, and Qdrant, with Redis-backed background jobs.

Stack, status, evidence, and public actions are rendered from the typed project record.

Category

AI Systems

Type

Platform

Priority

Flagship

Overview

What this project is

Aletheia runs keyword and vector retrieval side by side, reranks the merged results, and measures quality with evaluation pipelines and observability tooling. FastAPI services over PostgreSQL, OpenSearch, and Qdrant, with Redis-backed background jobs.

Information retrievalVector searchRAGSearch evaluationBackend systems

Problem

Why it matters

Retrieval systems fail quietly. Without evaluation loops and observability there is no way to tell why a query returned weak results or whether a ranking change actually helped.

Solution

Approach

One platform that serves hybrid retrieval and reranking behind an API, scores results with evaluation pipelines, and exposes ranking behavior for inspection.

Architecture

System shape and stack

Architecture

Retrieval quality as an observable system

The project is framed around retrieval, reranking, evaluation, and operational visibility rather than a single chat surface.

System Shape

Backend, index, and evaluation loops

The stack points toward API services, database state, background queues, vector search, keyword search, and evaluation workflows working together.

  • Python
  • FastAPI
  • PostgreSQL
  • Redis/RQ
  • OpenSearch
  • Qdrant
  • SQLAlchemy
  • Alembic
  • Docker
  • Next.js

Technical Highlights

Visible technical signal

  • Keyword and vector retrieval merged with reranking
  • Evaluation pipelines that score retrieval quality
  • Observability into how rankings are produced

What It Proves

Builder signal

Search infrastructure engineering: index design, ranking pipelines, evaluation methodology, and the backend services that hold them together.