About

Manar trains and deploys artificial intelligence on the hardware of the organization that will use it. Not on a shared cloud. Not on a server in another country. On a machine inside the building where the data already lives.

Every organization in Pakistan produces information it does not get to use. A factory records more than anyone reviews. A hospital keeps years of scans and files. A warehouse runs cameras nobody watches in real time. A school holds records going back decades. That information is accurate and current. It just sits there, unread, because reading it properly takes more time than anyone has.

Cloud AI asks an organization to send that information elsewhere before anything can be done with it. For a lot of organizations in Pakistan, that is not a minor request. A bank cannot send account records abroad. A hospital cannot send patient scans to a server it does not control. A factory does not want its production floor footage leaving the floor it was recorded on. Sovereignty is not a preference for these organizations. It is a requirement.

Start with a strong, openly available foundation model. Train it further on the specific conditions of one site. Run it on hardware sized for that site, kept inside the building it serves. The organization keeps its data. It keeps the hardware. It keeps the trained model. None of it depends on a connection to somewhere else.

The model and the hardware belong to you from the moment installation completes. Not licensed back, not rented, not tied to a subscription that lapses if a vendor stops operating. Owned.

This means the usual risk of working with a small or young vendor, that the vendor's continuity becomes your continuity, does not apply here the way it applies to a subscription product or a cloud service. If you ever want to run the model in house, retrain it yourselves, or bring in someone else entirely, nothing at your site depends on Manar continuing to exist to make that possible.

The engineering is direct rather than layered. The people who train and install a model are the same people who support it afterward. There is no account manager standing between you and the person who wrote the code.

The team is small by design. Every model that gets deployed is trained and reviewed directly, not passed to a support queue. That will not scale the way a subscription product scales, and it is not meant to. It is meant to produce something that works on one real site before it gets called finished, and something the client actually owns once it is installed.

Manar works between Spain and Pakistan, with its attention on the Pakistani market first. Textile and sportswear factories in Faisalabad and Sialkot. Cold chains carrying dairy and mango exports. Farms across Punjab. Hospitals, clinics, schools, and the government and finance offices that keep the country's records moving. Growth into the wider Gulf region will follow the same rule everything else here follows. Prove it on one site before calling it ready for the next.