CodeNinja, an artificial intelligence lab that builds self-improving systems, and Paranine, a company that serves open-weight AI models on hardware inside Pakistan, have announced a technology integration and commercial partnership. The announcement was made at CodeNinja’s office in Lahore. The two companies say the tie-up will deliver Sovereign AI Operating Systems for organizations in Pakistan.
The partnership aims to give institutions more control over three things: where their AI workloads run, how organizational knowledge is used and kept, and how the systems that make decisions on their behalf are governed. It combines CodeNinja’s AI applications, organizational ontology (a structured map of an organization’s systems, knowledge and relationships) and agentic systems, which are AI agents that carry out tasks, with Paranine’s in-country inference infrastructure. Inference is the step where a trained model produces answers, so running it inside the country keeps that processing within national jurisdiction.
Why location and control matter
The companies argue that sectors such as government, banking and healthcare need more than a capable model. They also need control over the infrastructure running it, the data feeding its outputs and the systems that apply its intelligence. Many organizations still reach AI through services hosted outside their national borders, which the companies say can raise questions about data residency, regulatory oversight and long-term dependence on critical systems.
In the companies’ description, Sovereign AI rests on two parts. The models must operate within the country’s jurisdiction, while the applications and intelligence built around them must stay under the organization’s own governance. CodeNinja already runs frontier open-weight models on its own Paranine instance, which the companies say gives the integration an existing foundation.
How the system is built
The operating system is built on Hyper Anthologies, CodeNinja’s ecosystem for engineering self-improving systems, and runs on Paranine’s inference infrastructure in Pakistan. It brings together four parts. Hyper Ontology builds a governed model of the organization so applications and agents share the same context. Hyper Pragma is the execution layer that lets agents work inside the organization’s environment, and Hyper Engram keeps persistent decision memory. Hyper Noesis applies interpretability, which lets teams inspect the concepts open-weight models rely on and spot when their judgment begins to drift.
Deployments can run on dedicated infrastructure in the country, within a customer’s own environment, or in air-gapped setups that have no connection to outside networks, where required. Open-weight models can be chosen for each deployment and replaced as stronger ones appear, without rebuilding the surrounding system. The companies say the systems and knowledge built for each customer can be transferred to that customer, so no single model provider becomes a point of dependence.
Mukhtar Ahmad, CEO and co-founder of CodeNinja, said that sovereignty depends on who controls “the systems, knowledge and decisions built around” AI, not only on where it runs. The announcement did not give a launch date or a list of first clients, and the partnership’s initial focus was not detailed in the material available.
Aasan alfaaz mein
CodeNinja aur Paranine ne Pakistan mein Sovereign AI systems ke liye sath milkar kaam karne ka elaan kiya hai. Is ka maqsad hai ke AI models mulk ke andar chalein aur idaron ka data aur faisle unke apne control mein rahein. Aam logon ke liye yeh tabdeeli foran nazar nahi aayegi, lekin bank, hospital ya sarkari dafater mein AI kaise istemal hoga, is par asar ho sakta hai.
