Physical AI

Training machines to service
the electric world.

India's EV fleet is growing faster than its ability to maintain it. We're building AI that learns from human technicians to close that gap — starting with motor controllers and battery systems.

The Problem

India's EV transition is outpacing its service capacity.

1.24M
1.68M
1.97M
2.55M
3.2M
4.8M
10M
FY23
FY24
FY25
FY26
FY27
FY28
FY30
Actual
Projected
Latest (FY26)

India EV registrations, all segments, millions of units1

93%2
of EV service workers lack any formal EV training
8.5%3
EV penetration today vs. 80% target for 2-wheelers by 2030
<20K4
trained EV technicians added per year against a need of 100–200K by 2030

What We're Building

AI that learns how humans fix EVs.

01

A dataset of physical work

Every service and repair in our network is recorded on video. We annotate each step — component identification, fault diagnosis, repair sequence — to build a structured dataset of how skilled technicians work on EV systems in the field.

02

Models that assist, then automate

Near-term, these models become an AI assistant for new technicians — guiding them through motor controller replacements, battery diagnostics, and BLDC system repairs in real time. The long-term goal is robotic service.

03

Starting in the field, not a lab

We operate our own EV rental fleet in Delhi, giving us direct access to real-world service data at volume. The data flywheel starts on the ground, not in a simulation.