Kaivera is an AI operations partner for businesses in Southeast Asia. We find where AI pays off in a company's operations, build it inside the client's own stack, and keep it running. Below is the engagement we're currently delivering, and the kind of work we take on.
Recharge Xolutions Sdn. Bhd.
Sector EV charging infrastructure
Status In progress
Engagement began September 2026
Field engineers commission and service chargers across sites. At each job they photograph the electrical meters, then type the readings into the job form by hand — around ten photos and multiple number fields per job, done on a phone, often somewhere with no signal.
It is slow at the point of work and it is the step where mistakes enter the record. A digit transposed on site is found weeks later, in an office, by someone who wasn't there.
[Content slot — volumes, number of sites, engineers in the field, and current time per job, once confirmed with the client.]
A vision model that reads the numbers off the engineer's photos and pre-fills the fields, so the job changes from typing to confirming. It is built into the client's existing app and cloud stack rather than sitting beside it as another tool to open.
The engineer still signs off on every reading. The model's job is to remove the keyboard, not the judgement.
image slot — capture screen or extraction flow diagram, to be supplied
In phases, with a feasibility gate before any committed build work. The client finds out whether the model reads their meters accurately enough before paying for a build around it.
Real photographs from real sites, tested against the model. Accuracy measured on the meter types actually in the field, including the badly lit ones.
A written go or no-go against the accuracy threshold agreed up front. No further build is committed until this is passed.
Integration into the existing app, then a trial with a small group of engineers on live jobs before it goes to the whole field team.
[Content slot — phase dates, accuracy threshold, and rollout scope, to be added as the engagement progresses.]
These are illustrative scenarios in the sectors we work in — the shape of problem we take on. They are not delivered projects and not client work.
Supplier price lists arrive as PDFs in a shared mailbox. A flow reads them, matches items to the catalogue, flags the changes that move margin, and writes the rest straight into the pricing sheet.
Each outlet's daily numbers land in one place overnight. The Friday report assembles itself, and the operations lead spends the morning on the two outlets that are off, not on the spreadsheet.
Job evidence captured on site is checked for completeness before the engineer leaves, so missing photographs are caught at the charger rather than at invoicing.