Back to all work
Content & Creative Automation

AI Content Agent.

An autonomous content production system connected to real marketplace inventory.

Generative AICreative workflowsAutomation

ai_content_agent.project

Concept illustration · not a product screenshot
From inventory to creative workflow.
A case study in progress

Experience and results reflect my résumé. Sections marked “Documentation pending” await a deeper walkthrough. Screenshots and decision records will be added when available.

PRODUCT SCOPE

A connected product experience.

  • AI content planning
  • Image & video generation
  • Real-product compositing
  • Automated creative workflows
  • Content quality evaluation
  • Scheduled publishing
  • Real marketplace inventory integration
01

Context

The AI Content Agent connects content planning, image and video generation, quality evaluation, and scheduled publishing in one production workflow, grounded in real marketplace inventory.

02

Customer problem

The intended workflow supports people managing marketplace content who need to turn available inventory into relevant creative. Specific user interviews and validation have not been supplied.

03

What I owned

Built an autonomous content pipeline that turns live marketplace data into social content, connecting automated planning, AI image generation, real-product compositing, quality scoring, and scheduled publishing.

04

Strategy & prioritization

Documentation pending

Document which content workflow was automated first and why. Explain how relevance to actual inventory and a usable review process shaped the scope before increasing publishing volume.

05

Constraints & options considered

Documentation pending

Describe the real boundary between autonomous generation and human review. What quality criteria gate publishing, and how does the workflow handle unsupported product claims or unavailable inventory?

06

Key decisions & why

Documentation pending

The decisions behind the first version, the alternatives, and the reasoning are still to be written up.

07

Execution

Built with TypeScript, the OpenAI API, and Claude Code. The pipeline combines live marketplace data, automated planning, AI image generation, real-product compositing, quality scoring, and scheduled publishing across TikTok, Instagram, and Facebook.

08

Outcome & learning

Ships six carousels and three videos daily across TikTok, Instagram, and Facebook, turning live marketplace inventory into scheduled social content.

09

Challenges & lessons

Documentation pending

Add an observed quality failure and the evaluation change it prompted. Distinguish a visually strong asset from one that accurately represents the underlying inventory.

10

What I’d change

Documentation pending

A proposed improvement is to define a review rubric for product accuracy, creative quality, and inventory freshness, then compare generated outputs against it. This has not been presented as completed testing.