What this is actually for
Your shop finds out a truck is broken when it's already out on a route with a crew on it.
What we built for ourselves
Nightly, on our own fleet: telematics, fault codes, work-order history and maintenance schedules, correlated into per-vehicle briefs.
Because inference is local, it runs the full history every night instead of sampling. That's what makes the ranking worth reading.
Hardest part wasn't the model — it was that three systems each named our trucks differently.
It arrives as one morning message, because nobody in a shop opens a dashboard.
What you get
- —A ranked morning list: what to look at first, and why
- —Per-issue diagnostic procedure, tools and parts required, and a time estimate
- —Vision grading of inspection and damage photos
- —Alert suppression for things the shop already actioned
Hear how it got built
We recorded the build of each of these, including the parts that didn't work.
- Episode 7 — The Fleet That Diagnoses Itselfrecording
Episodes publish as they're cut. Nothing here is a teaser for something that doesn't exist — the systems are already running.
What this doesn't do
It does not replace a technician. It hands one a prepared file before they pick up a wrench. And below roughly ten vehicles the economics don't work — the shop is usually outsourced and the brief has no reader.
What it runs on
Built on NVIDIA hardware and software we own and operate, not a reseller relationship.
Where it's going next: Abstracting ingestion so any telematics + work-order vendor combination maps in.
A member of the NVIDIA Inception program
Fifteen minutes, no deck.
Bring one specific problem in one sentence. We'll tell you whether it's worth building, roughly what it takes, and sometimes that the answer is don't build it.
Start hereTwo steps before we talk — it takes about five minutes.