Parts Intelligence

Overview

The medical-device service shop I work for runs on Tracmor — a legacy inventory app with no real API and two internal ID systems that disagree with each other. The Legacy Web Tooling kept day-to-day work moving; this is the data work meant to outlast it: making the shop’s own inventory trustworthy and queryable, and answering parts questions the legacy system never could.

What I built

• Reverse-engineering Tracmor’s data: nightly pulls of Tracmor’s SQL dumps into a queryable store, and reconciling its two disagreeing internal ID systems so that a part or an asset has one dependable identity.
• Asset deduplication: collapsing the duplicate and near-duplicate asset records that accumulated through the legacy system’s quirks into single canonical entries.
• A parts-compatibility store: a database of which parts fit and substitute for which, fused from shipment records, service-manual PDFs, and email archives, with per-source confidence scoring so that a fact’s reliability travels with it.
• A vision-LLM part identifier: a field tech photographs a part’s label and an on-device vision model identifies it, feeding the compatibility store. Still barebones, but already useful in the field.

Status

Early and ongoing. The reverse-engineering and deduplication carry forward into the shop’s proper inventory rebuild, an internal Tracmor replacement led by another engineer; the compatibility and part-identification work stands on its own. I pulled it out of the Legacy Web Tooling writeup so the durable data work isn’t buried behind a label printer.