Skip to content
FAHD MOHAMED
All work

Professional work · Multi-agent pipeline · FastAPI · Celery

AI SEO Agent System

An orchestrator and eight specialised agents that research, write, optimise, publish and track SEO content.

agents + orchestrator
8 + 1
agents + orchestrator
source: backend/agents/*

The problem

SEO content work is a chain of slow, manual steps: keyword research, search-result analysis, briefs, writing, speed audits, publishing and rank tracking.

My role

  • Engineered the system: multi-agent architecture, AI automation and backend development

Engineered during my role as AI & Software Engineer at Elenor Marketing Agency, Cairo (Jan 2026 – present). Screenshots come from sanitized demo environments: client and business data are replaced with demo data.

What it is

A FastAPI backend with Celery workers runs an orchestrator plus crawler, keyword, SERP, content, optimizer, speed, tracker and feedback agents. Keywords from Google Search Console are clustered with sentence-transformers. DataForSEO feeds search-result analysis, PageSpeed Insights feeds Core Web Vitals, and drafts are published to WordPress for human review. A weekly feedback loop re-weights the prompts based on ranking results.

Key features

  • 8 agents + orchestrator
  • Embedding-based keyword clustering
  • WordPress draft publishing
  • Daily rank tracking and weekly feedback loop
  • Next.js dashboard

Architecture

Sources

  • Search Console
  • DataForSEO
  • PageSpeed

Agents

  • Orchestrator
  • Keyword · SERP · Content
  • Optimizer · Speed
  • Tracker · Feedback

Workers

  • Celery + Redis
  • Flower

Stores & outputs

  • Postgres · ClickHouse · MinIO
  • WordPress drafts

AIOrchestrated specialist agents

AIEmbeddings for clustering

AIFeedback loop on ranking outcomes

Stack

Backend

Python 3.12FastAPICeleryRedisSQLAlchemy

Data

PostgreSQLClickHouseMinIO

AI & APIs

OpenRoutersentence-transformersDataForSEOSearch ConsolePageSpeedWordPress REST

Infra

Docker ComposeFlower

Engineering challenges

01

Closing the loop

Instead of generating content and moving on, a tracker agent records rankings daily and a feedback agent adjusts prompt weights weekly.

Not run on this PC: it needs Docker (Postgres, Redis, ClickHouse, MinIO, Celery), which is not installed. This case study is based on the source code only.

No screenshots for this project. See the architecture above.

Next project

Agent X