Sultan Meghji Announces Kundi 3: A Self-Improving Sovereign AI Platform That Institutions Own Outright
(MENAFN- EIN Presswire)
Frontier Foundry is a sovereign frontier lab for applied AI. It develops its own language models and the full stack around them: orchestration, an algorithm library, and a data lakehouse. It delivers all of it, weights included, to run in a customer's cloud account, on premises, or air-gapped. Kundi is not a wrapper around a third-party model. It requires no hosted model, API, or service from any other frontier lab, and the institution owns the models, the weights, and every improvement the system produces.
“Most of what is sold as sovereign AI is a rented model behind a firewall,” said Sultan Meghji.“If the weights belong to someone else and the system only gets better when that vendor ships an update, the institution doesn't control it. It leases it. We built Kundi so the institution owns the entire stack and the improvements accrue to them, not to us.”
Kundi 3 introduces recursive self-improvement across Kundi's three core systems. Each system's output becomes raw material for the others, so the platform's capabilities compound over time without leaving the institution's control.
● Conductor, the agentic layer. AI agents design, test, and refine the workflows and applications an institution runs, then execute them, keeping a verifiable record of every input, computation, model call, and decision. Where a workflow calls a language model, Conductor records the call and keeps it distinct from the deterministic computation around it.
● Cosmic, the research engine. Every night, Cosmic surveys new analytical methods, implements them, tests them, and adds those that pass to a library of more than 1,000 algorithms and models, with implementations in Python, Rust, C, C++, Ada, and COBOL, including the languages that still run bank cores and defense systems. Each validated addition is immediately available to Conductor's agents as a building block.
● Tarn, the data platform. A full-scope data lakehouse and warehouse built as a sovereign replacement for Snowflake or Databricks, Tarn handles ingestion, storage, SQL analytics, documents, vectors, model records, scheduled jobs, and governance on infrastructure the institution controls. As regulatory guidance changes, Tarn drafts the corresponding policy updates for approval. Once approved, they are enforced automatically across access, filtering, masking, and permitted use, and every governed query carries evidence of the policies applied.
Kundi's analytical execution is deterministic. A fixed algorithm version, input, and configuration produces the same result on every run, verified across the library's language implementations, and Frontier Foundry carries that property down to its own hardware for customers who want it. Tarn can replay any governed query against the data snapshot and policy version in force at the time. A platform that improves itself is therefore no harder to audit than one that stands still.
The complete platform runs in a customer's cloud account, on premises, or fully air-gapped, with quantum-resilient protection across the suite. Improvement processes run inside the deployment or centrally at Frontier Foundry, depending on how each institution configures them.
“Regulators don't object to systems that learn. They object to systems nobody can explain,” said Meghji.“Kundi's agents can draft a new workflow, add an algorithm, or propose a policy change overnight, but every change carries a deterministic record of what changed and why, and anything that touches how data may be used waits for a person to approve it. That is what lets a bank or an agency run a self-improving system and still answer an examiner's questions the next morning.”
Kundi 3 is live with customers and available now. Institutions interested in a walkthrough or a deployment assessment can contact Frontier Foundry at... or visit platforms/kundi/
About Sultan Meghji
Sultan Meghji is the founder and CEO of Frontier Foundry. He served as the inaugural Chief Innovation Officer of the U.S. Federal Deposit Insurance Corporation, and is a Non-resident Senior Technical Expert at the Center for Strategic and International Studies and a Visiting Fellow at George Mason University's National Security Institute. Meghji's writing on AI, finance, and national security has appeared in The Wall Street Journal and Bloomberg Opinion. He splits his time between St. Louis and Washington, DC.
More at
About Frontier Foundry
Frontier Foundry is a sovereign frontier lab for applied AI. It develops its own language models, algorithm library, orchestration, and data infrastructure and delivers them as Kundi, a full-stack agentic AI platform that regulated institutions own and run inside their own infrastructure, with deterministic analytics and quantum-resilient protection across the suite. Frontier Foundry's systems are in production at dozens of financial services firms and various agencies of the United States government. The company has SOC 2 Type II certification and is headquartered in the Washington, DC area.
Learn more at
Sultan Meghji, founder and CEO of Frontier Foundry, announced Kundi 3, the company's self-improving sovereign AI platform.
Frontier Foundry's agentic platform now builds its own workflows, expands its algorithm library nightly, and keeps data policy current with regulation.
If the weights belong to someone else and the system only gets better when that vendor ships an update, the institution doesn't control it. It leases it.” - Sultan MeghjiWASHINGTON, DC, UNITED STATES, November 10, 2026 /EINPresswire / -- Sultan Meghji, founder and CEO of Frontier Foundry, today announced Kundi 3, the new release of the company's sovereign AI platform for banks, asset managers, and government agencies. With Kundi 3 the platform becomes a self-improving system: its AI agents build and refine the workflows an institution runs, its research engine adds tested algorithms to the platform every night, and its data platform drafts updated access policies as regulation changes. Every improvement executes deterministically, and no change to how data may be used takes effect without the institution's approval.Frontier Foundry is a sovereign frontier lab for applied AI. It develops its own language models and the full stack around them: orchestration, an algorithm library, and a data lakehouse. It delivers all of it, weights included, to run in a customer's cloud account, on premises, or air-gapped. Kundi is not a wrapper around a third-party model. It requires no hosted model, API, or service from any other frontier lab, and the institution owns the models, the weights, and every improvement the system produces.
“Most of what is sold as sovereign AI is a rented model behind a firewall,” said Sultan Meghji.“If the weights belong to someone else and the system only gets better when that vendor ships an update, the institution doesn't control it. It leases it. We built Kundi so the institution owns the entire stack and the improvements accrue to them, not to us.”
Kundi 3 introduces recursive self-improvement across Kundi's three core systems. Each system's output becomes raw material for the others, so the platform's capabilities compound over time without leaving the institution's control.
● Conductor, the agentic layer. AI agents design, test, and refine the workflows and applications an institution runs, then execute them, keeping a verifiable record of every input, computation, model call, and decision. Where a workflow calls a language model, Conductor records the call and keeps it distinct from the deterministic computation around it.
● Cosmic, the research engine. Every night, Cosmic surveys new analytical methods, implements them, tests them, and adds those that pass to a library of more than 1,000 algorithms and models, with implementations in Python, Rust, C, C++, Ada, and COBOL, including the languages that still run bank cores and defense systems. Each validated addition is immediately available to Conductor's agents as a building block.
● Tarn, the data platform. A full-scope data lakehouse and warehouse built as a sovereign replacement for Snowflake or Databricks, Tarn handles ingestion, storage, SQL analytics, documents, vectors, model records, scheduled jobs, and governance on infrastructure the institution controls. As regulatory guidance changes, Tarn drafts the corresponding policy updates for approval. Once approved, they are enforced automatically across access, filtering, masking, and permitted use, and every governed query carries evidence of the policies applied.
Kundi's analytical execution is deterministic. A fixed algorithm version, input, and configuration produces the same result on every run, verified across the library's language implementations, and Frontier Foundry carries that property down to its own hardware for customers who want it. Tarn can replay any governed query against the data snapshot and policy version in force at the time. A platform that improves itself is therefore no harder to audit than one that stands still.
The complete platform runs in a customer's cloud account, on premises, or fully air-gapped, with quantum-resilient protection across the suite. Improvement processes run inside the deployment or centrally at Frontier Foundry, depending on how each institution configures them.
“Regulators don't object to systems that learn. They object to systems nobody can explain,” said Meghji.“Kundi's agents can draft a new workflow, add an algorithm, or propose a policy change overnight, but every change carries a deterministic record of what changed and why, and anything that touches how data may be used waits for a person to approve it. That is what lets a bank or an agency run a self-improving system and still answer an examiner's questions the next morning.”
Kundi 3 is live with customers and available now. Institutions interested in a walkthrough or a deployment assessment can contact Frontier Foundry at... or visit platforms/kundi/
About Sultan Meghji
Sultan Meghji is the founder and CEO of Frontier Foundry. He served as the inaugural Chief Innovation Officer of the U.S. Federal Deposit Insurance Corporation, and is a Non-resident Senior Technical Expert at the Center for Strategic and International Studies and a Visiting Fellow at George Mason University's National Security Institute. Meghji's writing on AI, finance, and national security has appeared in The Wall Street Journal and Bloomberg Opinion. He splits his time between St. Louis and Washington, DC.
More at
About Frontier Foundry
Frontier Foundry is a sovereign frontier lab for applied AI. It develops its own language models, algorithm library, orchestration, and data infrastructure and delivers them as Kundi, a full-stack agentic AI platform that regulated institutions own and run inside their own infrastructure, with deterministic analytics and quantum-resilient protection across the suite. Frontier Foundry's systems are in production at dozens of financial services firms and various agencies of the United States government. The company has SOC 2 Type II certification and is headquartered in the Washington, DC area.
Learn more at
Media Team
Frontier Foundry
...
Visit us on social media:
LinkedIn
Legal Disclaimer:
EIN Presswire provides this news content "as is" without warranty of any kind. We do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the author above.
You just read:
Sultan Meghji Announces Kundi 3: A Self-Improving Sovereign AI Platform That Institutions Own Outright News Provided By Frontier Foundry November 10, 2026, 12:41 GMT Share This Article Distribution channels: Banking, Finance & Investment Industry, Business & Economy, IT Industry, Science, Technology
EIN Presswire's priority is author transparency. We do our best to weed out false and misleading content. The content above is
the sole responsibility of the author who makes it available. If you have any complaints, kindly contact the author above.
Legal Disclaimer:
MENAFN provides the
information “as is” without warranty of any kind. We do not accept any
responsibility or liability for the accuracy, content, images, videos,
licenses, completeness, legality, or reliability of the information
contained in this article. If you have any complaints or copyright issues
related to this article, kindly contact the provider above.
Comments
No comment