Skip to content
← All work
01Flagship2026

BFL AI Studio

An AI-assisted e-commerce image-production platform structuring product upload, generation, review, approval and delivery in one workflow.

Role

  • Product Owner
  • Independent Builder

Discipline

  • AI Product Development
  • Creative Operations

Year

2026

BFL AI Studio — AI image production, from product selection through generation, QC review, model management and analytics.

01

Business context

BFL AI Studio removes the manual effort, inconsistency and coordination involved in producing e-commerce product imagery at scale. The platform brings product upload, model selection, generation, review, approval and final delivery into one structured workflow.

02

The challenge

Translate a repetitive, multi-stage creative-production problem into a controlled software workflow — without losing the quality judgement that makes e-commerce imagery usable.

03

Role & responsibilities

I conceived the product, defined the workflow, designed the user experience, structured the production logic and built the application. Stakeholder input validates operational requirements, while product definition and build ownership remain with me.

The platform is independently owned and built by me, with workflow input and validation from relevant creative, styling and operational stakeholders.

  • Defined the original product concept.
  • Designed user roles and permissions.
  • Designed end-to-end production workflows.
  • Designed interface and interaction logic.
  • Connected creative requirements with application logic.
  • Developed the application functionality independently.
  • Built batch uploads and scheduling.
  • Designed quality-control and approval stages.
  • Structured production-gallery and reporting requirements.
  • Coordinated creative, styling and operational expectations.

04

Strategic approach

  1. 01Start from the production line, not the interface — map every stage an image passes through before it is publishable.
  2. 02Make quality a gate rather than an afterthought: confidence review, approval/rejection and stylist QC are first-class steps.
  3. 03Keep models and backgrounds fixed systems so output stays consistent at volume.
  4. 04Design for batch reality — uploads, scheduling and bulk production rather than one-off generation.
  5. 05Give leadership visibility: production-line status, cost and time estimation.

05

Team & stakeholders

  • Product ownership and development carried independently.
  • Workflow input and validation from creative, styling, e-commerce operations and production stakeholders, plus technical or security stakeholders where required.

06

Solution & capabilities

  • UPC-based product records
  • Batch uploads
  • Fixed model and background systems
  • Hero and supporting product angles
  • Confidence and quality-control stages
  • Approval and rejection logic
  • Bulk production
  • Scheduling
  • Role-based access
  • Production-line visibility
  • Cost and time estimation

—

Technology & data

Stack

  • Modern web application stack
  • TypeScript-based interface
  • Supabase-backed data layer
  • AI image-generation workflow integrations
  • Structured role and workflow logic
  • Built with Antigravity, Claude Code and Codex
  • ComfyUI-based generative-image workflows

07

Process & workflow

  1. 01Product Upload
  2. 02Product Record / UPC
  3. 03Model and Background Selection
  4. 04Hero Image Generation
  5. 05Confidence / Quality Review
  6. 06Approve or Reject
  7. 07Supporting Angle Generation
  8. 08Stylist Quality Control
  9. 09Production Gallery
  10. 10Final Delivery

08

Outcome

  • Brings product upload, generation, review, approval and delivery into a single tracked workflow instead of scattered manual coordination.
  • Holds models and backgrounds as fixed systems, so imagery stays visually consistent across large product volumes.
  • Treats quality as a gate rather than an afterthought: confidence review, approve/reject and stylist QC each sit in the production line.
  • Handles batch reality — bulk uploads, grouped generation and scheduling rather than one-off image requests.
  • Gives leadership production-line visibility, including cost and time estimation.
  • Demonstrates product ownership, workflow design and the translation of creative-production problems into software systems.

09

Gallery

Dashboard showing the production review for a selected period: UPCs shot, photographs captured, AI images produced, the live versus non-live split, and the queues awaiting editor and stylist review.
Production History listing past generation sessions with per-product status, next to the approved hero image and its review trail for the selected product.
New hero screen: a product and an avatar are chosen as references, with an optional prompt, and generated against a fixed White Studio background.
Stylist QC station with generated lifestyle, full-shot, side and back frames grouped per product in folder batches for review and rating.
Angles QC showing individual hero images across products, each tagged with which angles have been captured and selectable for bulk action.
Model repository listing the AI human models available for generation, each recorded with gender, build and fit attributes.
Analytics view reporting generation volume, approval and first-pass rates, confidence distribution, status breakdown, per-model usage and top rejection reasons.