Case study
Internal toolAI Product Visuals Pipeline
A skill-driven pipeline that turns a client's chat request into an on-brand AI product photo, picking the right account, model and reference automatically.
4
brands on one pipeline, each with its own brand rules
Source: UBA-BRAIN client files · 29 Sep 2026
24
product drafts across 2 client jobs
Source: UBA-BRAIN job folders · 26–29 Sep 2026
FIG. 001 / ARCHITECTURE
How the system connects
Live model
Problem
Generating on-brand AI product photography for several cosmetics clients at once meant manually tracking which account and image model each client used, finding the right reference image, writing prompts that preserved exact packaging text, and filing outputs consistently — easy to get wrong under volume.
System
A chain of skills reads each client's account and brand rules, opens a uniquely-coded job folder, pulls the correct single product reference image, builds a prompt that locks in the real packaging text and photorealism requirements, then generates through either the Magnific REST API (Nano Banana Pro / GPT Image 2) or the Higgsfield CLI depending on which account that client owns. Every draft is auto-named and logged, and approved images are copied into a final folder and back into the product's reference set.
FIG. 002 / HOW IT WORKS
How it works
This pipeline is the automation layer behind the design side of the business: a set of chained AI-agent skills that take a plain-language image request and produce a correctly branded, on-model product photo without the operator touching an API console. It enforces per-client credit and account isolation, never lets a competitor's photo leak in as a style reference, and keeps every draft and approved final traceable to the exact prompt and settings that produced it — letting one designer run AI photo production for multiple brands in parallel.
Stack
What it runs on
Related systems
UBA-BRAIN
Internal toolA shared markdown knowledge base and rule system that lets any AI CLI on any machine run the same client work with the same rules, memory and skills.
TruckPulse
Internal toolAutomated load-profitability and dispatcher-performance monitoring for a trucking company, built from Gmail broker PDFs and Telegram driver-rate messages.
Upwork Command Center
Internal toolA personal live dashboard that streams new Upwork jobs every 30 seconds, auto-translates and summarizes them, and drafts human-sounding proposals on demand.
Want a system like this for your team?
We start by mapping one workflow on a 30-minute call.