The RegenITAD AI Asset Listing Pipeline turns photos, asset labels, and operator notes into normalized, reviewed, marketplace-ready listing drafts. Built for the ITAD, recommerce, and surplus IT equipment world — where every server, switch, and workstation arrives in a slightly different state and has to be listed accurately without burning a human on each one.
IT asset disposition and recommerce are bottlenecked at the listing desk. Each unit has factory specs, current configuration, photos, condition notes, and a service tag that may or may not still be legible. Producing a clean marketplace listing from that mess is slow, repetitive, and expensive — and inconsistent listings cost real money on resale.
The Listing Engine attacks the slow part: extraction, normalization, and draft synthesis. It does not attack the judgment part — a human still approves every draft before it can be exported to eBay, the RegenITAD storefront, or a CSV pipeline.
Seven explicit stages. The AI handles extraction and synthesis; the accountable human handles judgment, exception review, and final action.
End-to-end story
Asset photos / labels / notes → OCR / service tag extraction → Dell/vendor enrichment → factory-vs-current configuration review → listing draft generation → human review → export to eBay / RegenITAD storefront / CSV.
Photos of the chassis, asset labels, and operator notes. Sample / synthetic inputs only — no real customer asset tags.
Optical character recognition pulls service tag / model / serial candidates out of the label photos and notes.
Tag candidates are looked up against vendor specs (e.g. Dell) to recover the factory configuration and product family.
Factory-vs-current comparison highlights what was changed: RAM, drives, CPU, GPU, daughter cards — the listing differentiators.
AI composes a normalized listing draft: title, condition, spec table, photos selection, and suggested price band.
An operator approves, edits, or rejects the draft before anything leaves the system. Nothing publishes without sign-off.
Approved drafts can be exported to eBay, the RegenITAD storefront, or CSV for downstream tools. Export, not auto-publish.
The AI is doing extraction, lookup, normalization, and draft synthesis — not pricing decisions, not publishing, not selling.
Image preprocessing plus OCR to lift text off asset labels and operator notes, with confidence thresholds on each field.
Service tag / part number candidates are normalized and resolved against vendor metadata to attach a factory spec.
Structured comparison of factory-vs-current configuration so the listing reflects reality, not a stale datasheet.
Templated draft generation that pulls from the normalized spec, the diff, and the operator notes — review-ready, not publish-ready.
Planned MVP — what the system will and will not do.
The listing-engine work is tracked in the private RegenITAD implementation repository. The page you are reading is the public proof surface; the tracker is where the work itself happens. Tracker access is private — reviewers can request access via the demo-access form.
Request tracker accessWhen the protected MVP is ready, allow-listed reviewers will be able to step through the pipeline on sample / synthetic assets. Register interest now — operator-reviewed, out-of-band approval, no marketing list.
I take on ITAD, recommerce, ops-heavy listing pipelines, and AI-assisted document generation through scoped work with concrete artifacts.