Senior Rails Developer (Contract)
Manheim · Nov 2015 – Apr 2016
Built SFAPI, a greenfield Rails service brokering post-auction paperwork between Manheim's online platforms and rate-limited AS400 legacy systems — accept fast, queue durably, retry safely. It was still in service six years later.
Manheim runs the world’s largest wholesale vehicle auctions, and this was my second tour with a Cox Automotive company — DealerMatch, a startup wholly owned by Cox, had been the first. One of Manheim’s enterprise architects recruited me directly; I came in through a staffing agency by choice, as the most senior engineer on a team of eight: four developers, a dev-ops engineer, QA, a project manager, and the architect.
After the gavel
When the gavel bangs on the last vehicle of an auction, the sale isn’t done — titles, paperwork, and post-close details all have to be buttoned up before money moves. Those processes lived on legacy AS400 systems that scaled poorly and enforced strict rate limits, while Manheim’s fast-growing online portfolio needed to talk to them constantly. Something had to play mediator.
That something was the Sale Facilitation API (SFAPI), which I built
greenfield in Ruby on Rails. Its core design bet: when the caller can’t do
anything about an upstream failure, returning an error is pointless. So SFAPI
validated incoming work, acknowledged it immediately with a 202 Accepted, and
enshrined it in durable storage before handing it to Sidekiq for asynchronous
processing. Temperamental upstream failures retried themselves; jobs that truly
failed could be triaged and replayed later — by a human if necessary — after the
underlying problem was fixed. The front-end stayed snappy, transaction errors
stopped reaching callers, and the customer experience never took the hit for an
AS400 having a bad day.
Greenfield, done right
Because SFAPI was new construction, we could install modern practices from the first commit. Manheim still largely lived in a physical data center of racked servers, with AWS just beginning to spread internally — so we built the service on twelve-factor principles (a natural fit for disposable virtual servers, and an approach my Heroku experience had already proven out), with fully automated testing and delivery replacing manual release processes, and AWS OpsWorks automating instance builds and deployments at a time when hand-tended, long-lived EC2 servers were still the norm.
The long tail
SFAPI outlived my contract by a wide margin. When I returned to Cox Automotive in 2018, it was still in service — and as late as 2021 the team that inherited it was still pinging me for guidance. The relationships lasted too: the architect who recruited me for this contract rose to Senior Director of Software Engineering — and he’s the one who brought me back to Cox in 2018, where the team’s dev-ops engineer was once again a teammate. Some engagements end; the good ones keep hiring you back.