A truck comes into the workshop with a fault nobody can place. Three technicians look at it, and each of them has a theory. Two hours later they are still guessing, the bay is occupied, and the customer has started to ask questions that nobody wants to answer.
What makes this frustrating is that the answer was in the building the whole time. It was in the bus traffic the vehicle has been generating since it was started. It was in the service history of the eleven other units of the same model that came through last year. It was in what the senior engineer wrote in the job card in 2027, in his own shorthand, which nobody has read since.
None of that is missing. It is just not in one place, and none of it is in a form anyone can consult while standing next to a vehicle with a torch in one hand.
What VEDA does
VEDA is a small language model that reads what the machine is actually saying, and it runs on a box inside your workshop. It takes the signals on the bus, the pattern of past repairs on that model, and what previous engineers wrote — and it comes back with a ranked list. Check this first. Then this. And here is why.
Two things about that list matter more than the list itself.
The first is that it is ranked, with reasons, not a single confident answer. A diagnostic that says "it is the injector" and is wrong twice teaches the workshop to ignore it. One that says "start here, because these three things are true" survives being wrong, because the technician can see the reasoning and disagree with it.
The second is that the disagreement is the point. When the technician says no, that answer goes back into the record. It is the only thing VEDA learns from — not scraped forums, not the OEM's optimistic manual, but the verdicts of the people who actually opened the machine. Over a year, that turns three technicians' instinct into something the fourth one can use on their first day.
The knowledge is already in your building. It has just never been written down in a form a machine could read.
Why it runs on your hardware
A large general-purpose model knows a little about everything, needs a datacentre, and cannot help you when the uplink is down — which, in a workshop, it often is. A small model trained on the machines you actually own knows a great deal about those twelve things, fits on a box you can buy and cool, and answers in about the time it takes to look up from the engine.
It also never leaves. The bus traffic, the job cards, the photographs, the verdicts — all of it stays on the machine in the corner of your building. That is not a setting we offer. It is the reason the system is shaped the way it is, because a record of how your fleet fails is worth something, and it should be worth it to you.
What you end up with
On day one, a technician gets a better place to start than a theory. That alone pays for the exercise.
What accumulates is more interesting. After a year, the workshop has a written record of its own diagnostic reasoning: what went wrong, what was tried, what it actually turned out to be, and who confirmed it. When someone retires, that does not leave with them. When a new engineer starts, they are not beginning from zero. And when the OEM asks why this model keeps coming back, you have something to show them.
Every step of it ends with a person. VEDA proposes; the technician decides. That is the one part of this we are not trying to automate.
VEDA is one of three products in VarahiEdge, our OEM engineering AI division — alongside Nirdesh, which replaces the machine manual, and VCAD, which generates the part. All three run on the same box, and write into the same record.
Read about VarahiEdge →