Skip to content
All work

TSL (client: Petco)

Petco Depot Operations Platform

An integration and workflow layer for Petco's device depot: JSM, Jira Assets, Fishbowl, and carrier APIs behind role-based operational interfaces.

Aug 2026 – Present

One orchestration layer over four systems that each own their own truth

As forward-deployed engineer I'm designing the integration layer for Petco's device depot, mid-transition off its incumbent vendor. JSM, Jira Assets and Fishbowl each stay authoritative over what they own; the binding constraint is serial-level chain of custody, so the migration quarantines unknown records rather than guessing. A wrong serial that looks right is worse than one flagged for reconciliation.

4
systems of record orchestrated
Serial-level
chain of custody
98%
contract CPI target

The problem

Depot operations for a national retailer span Jira Service Management, Jira Assets, Fishbowl Inventory, and carrier systems, mid-transition from an incumbent depot vendor. Every role currently has to understand every system, and serial-level custody evidence lives in no single place.

What I built

  • 01Designing an orchestration layer over the confirmed systems of record rather than a replacement platform, so each source keeps authority over the data it owns.
  • 02Writing a system-of-record matrix per data domain (ticket, asset, inventory, shipment, configuration, identity, KPI, audit) to be confirmed with client engineering before synchronisation rules are implemented.
  • 03Building role-based interfaces so a store user sees a request and its status, an executive sees queue and CPI health, and a technician sees a simplified work queue with Start, Pause, Complete and Escalate.
  • 04Making transition routing configuration-driven, phased by SKU, asset class, store, request type and effective date, with authorised manual override.
  • 05Staging the legacy migration as raw copy, map, normalise, deduplicate, classify and validate, so only reconciled records reach the operational path.
  • 06Engineering the integration for failure: idempotent writes, retry with backoff, a visible failed-transaction queue, dead-letter recovery, and per-connector health.
  • 07Capturing audit events that carry user, timestamp, previous and new state, asset and ticket, custody point and source system.
  • 08Establishing CPI data lineage for the three contract measures so every reported number traces to approved source data and documented exclusion rules.

Outcome

Contract executed and architecture in technical discovery with the client's engineering stakeholders, working to a 21 to 30 day initial-structure milestone and a 45 to 60 day controlled-transition window, with three contract CPIs (standard fulfillment, inventory accuracy, device-configuration accuracy) targeted at 98%.