Two Pakistani technology companies have joined hands to give local organisations what they have long wanted: artificial intelligence that runs inside the country and stays under their own control. CodeNinja, an AI lab building self-improving systems, and Paranine, an infrastructure company serving open-weight models on hardware in Pakistan, announced a technology and commercial partnership at CodeNinja’s office in Lahore.
The deal brings together two halves of the same idea. CodeNinja contributes the software layer, the applications and AI agents that carry out tasks for an institution. Paranine contributes the computing infrastructure inside Pakistan on which those models run. Together they will offer what they call Sovereign AI Operating Systems, for organisations that cannot have their most sensitive work processed on servers abroad.
Controlling Where the Intelligence Runs
For organisations in government, banking, healthcare and other regulated fields, bringing AI into daily work involves far more than picking a clever model. Someone has to decide where that model lives, what data feeds its answers, and how its intelligence is put to use. Get those answers wrong, and the organisation can break rules it never meant to break, or hand its knowledge to a system it cannot oversee.
Many Pakistani organisations still reach their AI through services hosted outside the country’s borders, where data sits in another jurisdiction and regulators struggle to inspect decisions. Genuine sovereign AI needs two things at once: the machines running the models must sit within the country’s jurisdiction, and the applications built around them must stay under the organisation’s own governance.
Built on Sovereign Local Infrastructure
The combined system is built on Hyper Anthologies, CodeNinja’s framework for systems that improve with use, and runs on Paranine’s inference infrastructure in Pakistan. CodeNinja already runs leading open-weight models on its own Paranine instance, so the integration has a working foundation.
The architecture rests on four capabilities. Hyper Ontology builds a governed picture of the organisation, linking its systems and knowledge so applications and agents share one view. Hyper Pragma is the layer where AI agents do the work, inside the organisation’s own environment. Hyper Engram gives the system a lasting memory of its decisions, so the next job is done better. Hyper Noesis brings interpretability to the models the organisation owns, letting teams see what a model relies on and when its judgement drifts.
Because the system uses open-weight models, an organisation can choose the model for each deployment and swap in a stronger one when it appears, without rebuilding everything. Deployments can run on dedicated infrastructure inside the country, in the customer’s own environment, or fully air-gapped. When the work is done, the systems and intelligence built for that deployment are handed to the customer, so the knowledge stays with the institution.
“Sovereignty is determined not only by where intelligence runs, but by who controls the systems, knowledge and decisions built around it. Our partnership with Paranine gives institutions in Pakistan the ability to deploy AI on infrastructure within the country while building intelligence that remains under their own control,” said Mukhtar Ahmad, chief executive and co-founder of CodeNinja.
Built for Institutions That Need Control
The partnership begins with the sectors where control matters most: government departments, banks and financial services firms, hospitals and other fields where data governance is non-negotiable. These organisations have been slowest to adopt AI, not because they doubt its value, but because the available options asked them to give up too much control. CodeNinja’s Forward Deployed Engineers will work inside customer environments to build the applications and agents, while Paranine supplies the computing capacity.
Bridging Sovereign Infrastructure and Enterprise AI
Paranine presents itself as Pakistan’s first sovereign AI Inference Cloud, letting organisations run AI models on locally hosted hardware. Its platform supports open-weight models on Huawei Ascend processors and NVIDIA graphics cards, and offers an OpenAI-compatible gateway, dedicated inference endpoints, and billing in Pakistani rupees.
CodeNinja describes itself as a Middle Eastern-American AI lab, with more than 350 engineers across Dallas, Riyadh, Santiago and Lahore, and over 400 projects delivered.
“AI adoption in Pakistan will be driven by organizations moving beyond experimentation to real-world deployment. This requires both reliable, sovereign AI infrastructure and applications that address practical business needs. By combining Paranine’s inference capabilities with CodeNinja’s enterprise AI systems, we aim to make that transition easier, more accessible and scalable for organizations across the country,” said Ahad Khurshid, founder and chief executive of Paranine.
The announcement lands as Pakistan tries to keep its AI future in its own hands. The federal cabinet approved the country’s first National AI Policy on 30 July 2025, built on six pillars including data sovereignty and a national computing grid, aiming to train one million AI professionals by 2030. What has been missing is the layer turning that computing power into working systems inside real institutions, the gap this partnership claims to fill.
Whether it succeeds depends on what happens next. Regulated organisations will want real deployments, not announcements, and will judge it on reliability. Pakistan still has no complete personal data protection law, and the draft bill has waited for years. But the direction is clear: the companies building Pakistan’s AI infrastructure are no longer asking whether the country needs its own systems. They are building them.
