← Work

Dealerware

One month for the first brand. Days for the rest.

Replaced fragile Go and Python AWS services with one agnostic Databricks-based integration layer — the first brand took a month, every brand after took days.

1 classto add Nissan — not a new pipeline

Problem

Vehicle and user data sync ran on a scatter of Go and Python services on AWS, stitched together with SNS and SQS. It broke under new requirements, and adding an OEM brand meant significant rework rather than configuration — so every new partner carried the cost of the last one all over again.

Action

Designed an agnostic Databricks-based integration layer for the sync, consolidating those services into a unified Rails architecture instead of maintaining several runtimes for one job.

Built the first full brand implementation on top of it for Volvo, including infleet/defleet parking lot functionality — deliberately as the general case, not as Volvo’s case.

Result

The first implementation took roughly one month. Subsequent brands — Nissan, Audi and others — came down to days, through plug-and-play configuration. Nissan specifically needed one additional class: not a new pipeline, not a rewrite.

The parking lot functionality was later promoted from Volvo-specific to platform-wide, and became a feature actively used in Dealerware’s sales pitch to prospective OEM partners. Architecture that turned into something the company could sell.