Itemize

The premise

Inventory software is usually designed for warehouses. Itemize starts with a different environment: ordinary people trying to remember what they own, where it is, and how to find it again without turning their home into a cataloging project.

The system

Itemize is a full-stack, self-hostable inventory application. Its ingestion flow uses hand-built image segmentation and recognition pipelines to turn photos into useful inventory candidates. Semantic search helps retrieve items when the user remembers meaning or context rather than an exact label.

Why build the pipeline

The hard part is not placing a form in front of a database. It is reducing the effort required to create trustworthy records. Image processing can propose structure, but the product still needs to handle ambiguity, corrections, and the gap between model confidence and user intent.

Decisions in progress

• Keep the complete product self-hostable rather than depending on a managed service for its core value.
• Use vision to accelerate ingestion, not to remove user control.
• Make semantic retrieval complementary to structured inventory fields.
• Design around household behavior instead of importing warehouse workflows.

Current evidence

What I am learning

Applied AI becomes product engineering at the edges: uncertain input, useful defaults, correction loops, evaluation, and the operational cost of every model call.