Kanza AI’s Clinical Reasoning System Launches with Lightning AI on NVIDIA Infrastructure in the US
Kanza AI, ranked first in independent evaluations against Claude, GPT, Gemini, and OpenEvidence, has gone live at Freyja Clinic
SEATTLE, Wash. and SILICON VALLEY, Calif., June 08, 2026 (GLOBE NEWSWIRE) -- Lightning AI today announced that Kanza AI’s Clinical Reasoning System (CRS) is now live at Freyja Clinic in Redwood City, California, running on Lightning AI’s agentic AI platform, GraphN. The Kanza AI CRS is available now for healthcare institutions, clinics, and peer-nominated physicians.
This is the first production US deployment of Kanza’s medical AI built to reason through diagnoses alongside physicians at the point of care. GraphN is the execution layer beneath the CRS. It gives a healthcare institution a direct path from its own proprietary data to working clinical AI, running in whatever environment its compliance regime demands: cloud today, with sovereign — on-prem or air-gapped — deployments available.
Kanza’s edge is not just the model. It is the substrate. Kanza is built on longitudinal clinical data samples from more than 300 terabytes of proprietary data spanning a network of 90-plus hospitals with 400-plus locations, plus specialty clinics.
“The organizations that will define the next decade of enterprise AI are sitting on proprietary data that general-purpose infrastructure cannot touch,” said Saurabh Giri, Chief Product & Technology Officer, Lightning AI. “Kanza’s moat is its data access, knowledge graph, and reasoning layer. GraphN gives them the execution layer to take it all to production at scale. This combination is what makes it defensible.”
Unlike documentation tools or medical search engines, the Kanza AI system works a case the way a physician does — weighing evidence, flagging where it’s uncertain, and citing every step so its reasoning is auditable and reproducible.
In an independent evaluation of 1,060 randomly sampled cases, scored across 32 dimensions of clinical reasoning, it placed first on every dimension against Claude, GPT, Gemini, and OpenEvidence. In blinded head-to-head scoring, three independent AI judges each ranked it first. In addition, a leading health system’s medical panel rated Kanza first after a nine-month evaluation that screened frontier LLMs and more than 200 AI startups.
On the MedXpertQA and USMLE benchmarks it posts the highest accuracy score of any model evaluated and the widest reasoning lead—roughly 0.8 points (out of 5) ahead of every frontier model tested—and its reasoning layer lifts every model it runs on, open-source and frontier alike, by an average of 0.6 points.
“It was designed for me as a physician and for my clinic. It reasons the way I do, to support me. A colleague, not a co-pilot or just a scribe. This is how medicine should work with AI,” said Dr. Jan Rydfors, MD, FACOG, a Stanford-trained OB-GYN, Assistant Clinical Professor at Stanford Medical Center, and co-author of “The Red Book,” a clinical reference for OB-GYNs. The CRS is in live use at Freyja Clinic for its physicians and nurse practitioners, deployed on Lightning AI’s NVIDIA compute infrastructure.
Kanza tunes best-in-class open-source models against its proprietary clinical substrate; GraphN orchestrates the models, tools, and agents behind Kanza’s workflows across cloud, on-prem, hybrid, and air-gapped deployments.
Healthcare systems hold decades of proprietary data, and researchers have built models that beat frontier labs on clinical benchmarks. That progress rarely reaches the practitioner, because four constraints converge: sensitive patient data cannot leave the institution; vendor lock-in blocks access; clinical environments cannot accept the performance tradeoffs of RAG-only systems; and enterprise compute cloud is costly to deploy at scale.
“Healthcare has a knowledge access problem—how to turn proprietary data into intelligence that works at the point of care,” said Samir Arora, Founder and CEO of Kanza AI. “GraphN, running on Lightning AI’s NVIDIA infrastructure, provides the production backbone that helps deliver Kanza’s CRS. Every clinical decision makes the system more grounded, more auditable, and more local.”
More than 30 physicians across leading US and Japanese institutions are co-developing the system through Kanza’s peer-nominated Specialists Program, with no paid acquisition.
About Lightning AI
Lightning AI is the company behind PyTorch Lightning and an NVIDIA Cloud Partner. GraphN is Lightning AI’s Agentic AI platform—a unified hardware and software stack that lets organizations build, evaluate, compare, and operationalize AI workflows on dedicated NVIDIA GPUs in complex enterprise environments, across cloud, hybrid, on-prem, and air-gapped deployments. Learn more at lightning.ai.
About Kanza AI
Kanza AI builds clinical reasoning AI for physicians at the point of care, co-developed with clinicians, trained on proprietary longitudinal clinical data and deployable across cloud and sovereign — on-prem or air-gapped — environments. The company is in live clinical use across institutions in the US, is a member of the NVIDIA Inception Program, and was selected by Google in 2026 as top-tier Scale AI in Health. Kanza was founded in Silicon Valley by Samir Arora, with co-founders Jeet Kaul, Sal Arora, Raj Narayan, and Dr. Toby Cosgrove, former CEO of Cleveland Clinic. Learn more at kanza.ai.
Media Contacts
Lightning AI: Lilly Savin | Lsavin@thisisoutcast.com
Kanza AI: Liz Larson | info@kanza.ai
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.
