Parinita Sovereign
Customer Data Never Leaves Jurisdiction. AI Comes to the Data.
Six classification zones. Four isolation layers. AI travels to your data — never the reverse.
Sovereign is the third pillar of the Parinita platform alongside AI Edge infrastructure and Nexus collaboration. One principle: customer data never leaves their jurisdiction. AI comes to the data, not the other way around. Six components enforce that at every layer.
AI Services Require Your Data to Leave Jurisdiction
Every major AI service requires sending your data to a vendor's cloud infrastructure — creating compliance exposure the vendor cannot legally fix.
HIPAA, GDPR, ITAR, and sovereign AI regulations all restrict where regulated data can travel. No cloud AI provider can make your data sovereign by contract alone — the architecture has to enforce it.
-
Regulated Data Cannot Go to Cloud AI
HIPAA PHI, ITAR-controlled data, and GDPR-regulated personal data cannot legally travel to most US cloud AI providers. The architecture has to enforce residency, not just the contract.
-
Consent Is a Checkbox, Not a System
Most consent management is a UI workflow with no enforceable data lifecycle. Purpose limitation, expiration, and deletion proof are not technical guarantees — they are claims.
-
Multi-Jurisdiction AI Is Operationally Impossible Today
Operating under GDPR, HIPAA, and ITAR simultaneously requires satisfying the intersection of all three — a constraint no cloud AI vendor can compute and enforce automatically.
Six components. One principle. Data stays. AI travels.
Sovereign's architecture enforces data residency at the legal, network, storage, and cryptographic layers simultaneously — not by policy, but by architecture.
The 101-LLC corporate structure encodes jurisdictional choice at the corporate level — a subpoena targets a specific LLC entity, not a multi-tenant vendor account.
-
Sovereign Data Vault
Encrypted storage with six classification zones — PUBLIC, INTERNAL, RESTRICTED, PHI/HIPAA, PCI/PCI-DSS, ITAR — using MinIO, AES-256-GCM, and HashiCorp Vault with per-POP keys.
-
Consent and Governance Engine
Granular consent, purpose limitation, and expiration backed by PostgreSQL and Casbin policy. Consent is a first-class artifact with an enforceable data lifecycle — not a checkbox.
-
Chrysalis Audit Trail
Every consent grant, residency event, and deletion proof anchors on-chain via Chrysalis VRC smart contracts. Audits become queries, not projects.
-
Sovereign AI Processing
Local inference, fine-tuning, and RAG via vLLM, Axolotl, and Milvus — the model travels to the data, never the reverse. AI comes to the data.
-
Insight Exchange
NVIDIA FLARE federated learning and OpenDP differential privacy let customers trade insights across organizational boundaries with cryptographic privacy guarantees. Insights move; data does not.
-
Legal Isolation Layer
Each tenant's data binds to a specific 101-LLC entity in the Parinita corporate structure. A subpoena targets that entity, not a multi-tenant vendor account.
-
Network Isolation Layer
Workload identity attached to every packet at the ConnectX-7 NIC via Crucible eBPF/XDP — network-level residency enforcement below the OS.
-
Storage Isolation Layer
Data partitioned per tenant on Plane 5, physically bound to the assigned jurisdiction POP. Data cannot be routed to a non-jurisdiction POP by architecture.
-
Cryptographic Isolation Layer
Every meaningful action signed with HSM-backed validator keys. Per-POP key management ensures cryptographic isolation even within shared hardware.
The Plane Model
Orchestra introduces "planes" — logical groupings of hardware optimized for a specific workload class. Unlike Kubernetes node pools, planes represent fundamentally different hardware architectures with different drivers, network requirements, and scheduling semantics.
The plane model is what makes Orchestra different from every other orchestration tool. Kubernetes sees nodes. Orchestra sees purpose-built hardware tiers and routes workloads accordingly.
Four independent layers. No single point of failure.
Data residency is enforced simultaneously at the legal layer (101-LLC entity binding), network layer (Crucible NIC identity), storage layer (Plane 5 jurisdiction-bound partitions), and cryptographic layer (HSM-backed signatures).
- 01Classification zoneSix zones — PUBLIC, INTERNAL, RESTRICTED, PHI/HIPAA, PCI/PCI-DSS, ITAR — each with independent encryption keys and access policies.
- 02Jurisdiction bindingData binds to a specific 101-LLC entity and a specific POP jurisdiction. Routing to a non-jurisdiction POP is blocked at the NIC by Crucible.
- 03Consent lifecyclePurpose, scope, and expiration defined at consent grant. Deletion proofs anchor on Chrysalis — enforceable by architecture, not just policy.
- 04AI processing modelvLLM, Axolotl, and Milvus run locally on Plane 5 at the jurisdiction POP. Models travel to data — data never travels to models.
- 05Federated learningNVIDIA FLARE enables cross-organizational model training with OpenDP differential privacy. Gradient updates move; underlying data does not.
- 06Multi-jurisdiction intersectionSovereign computes the intersection of all applicable compliance constraints automatically — the most restrictive rule governing each data element applies.
- 07Audit query modelEvery residency event and consent change anchors on Chrysalis via VRC smart contracts. Audits are Chrysalis queries — not log aggregation projects.
- 08Deletion proofDeletion events anchor on-chain with cryptographic proof of erasure — court-admissible evidence of GDPR Article 17 compliance.
import sovereign
client = sovereign.Client(seat_token='your_seat_token')
# List data classification zones for your tenant
for zone in client.vault.zones():
print(f"Zone: {zone.name}, Jurisdiction: {zone.jurisdiction}, POP: {zone.pop_id}")
# Run local inference — model comes to data
result = client.ai.infer(
model='parinita-gpt-7b',
prompt='Summarize this PHI record',
data_zone='PHI/HIPAA',
jurisdiction='us-east'
)
print(f"Result: {result.text}, Residency proof: {result.chrysalis_tx}")Proven at scale. Not in a lab.
Parinita AI Edge is the production deployment of the Parinita platform and the largest heterogeneous AI infrastructure deployment in the United States.
Parinita AI Edge
The most complex heterogeneous AI infrastructure in the United States. 101 sites, 9 planes, 12,000+ nodes, 4 accelerator vendors, dual network fabrics, four-layer tenant isolation — all through a single sovereign control plane.
Network & Security Infrastructure
- Multi-vendor acceleratorsFour accelerator vendors — Intel Habana, NVIDIA, AMD, Qualcomm — orchestrated through one control plane with unified scheduling, monitoring, and lifecycle management.
- Dual-fabric networkingCisco production fabric and Arista GPU backend fabric operating as a coordinated system, bridged by identity-aware routing.
- Nationwide scale101 sites across 42 U.S. states, each operating autonomously with a local control agent and a sovereign cross-site routing plane.
- Multi-tenant isolationFour-layer defense-in-depth: VXLAN VNIs, identity-routing, Palo Alto firewalls, and Cilium eBPF — validated across every plane and site.
- Compliance readinessFIPS 140-2 at launch, with FedRAMP Moderate, CJIS, and IL4/IL5 certification paths active through Parinita compliance profiles.
- Sub-millisecond routingEvery request classified and dispatched in under 1ms, enabling real-time SLA enforcement without perceptible overhead.
Built for regulated industries where data residency is a hard constraint.
Sovereign computes the intersection of multiple compliance constraints — GDPR, HIPAA, ITAR — and enforces the most restrictive data handling rule that satisfies all simultaneously.
Sovereign is the runtime layer that produces residency evidence. Choosing which controls apply is still a human judgment.
-
Healthcare and Life Sciences
HIPAA PHI must stay in jurisdiction. Sovereign AI Processing brings models to PHI — no PHI leaves the POP. Guardian agent provides PHI-safe summarization.
-
Financial Services
PCI-DSS and SEC 17a-4 data residency requirements met at the storage, network, and legal layers simultaneously. Ledger agent for KYC/AML with in-jurisdiction processing.
-
Defense and Government
ITAR-controlled technical data and classified material stay within ITAR-compliant entity boundaries. Atlas agent for classified document handling within jurisdiction.
-
Multi-Jurisdiction Enterprises
GDPR, HIPAA, and sovereign AI regulations simultaneously. Sovereign computes the intersection and enforces the most restrictive data handling rule for each data element.
-
Federated AI Research
Organizations that need to collaborate on AI model training across institutional boundaries without sharing underlying regulated datasets. FLARE + OpenDP makes insights portable, data stays put.
Deployment Models
Sovereign deploys as a platform service — data classification zones and consent policies configure at tenant onboarding. Chrysalis audit trail activates automatically.
Data classification zones and consent policies configure at tenant onboarding. Chrysalis audit trail activates automatically on first data write.
vLLM, Axolotl, and Milvus deploy at your jurisdiction POP. Inference, fine-tuning, and RAG all run locally — no data leaves.
NVIDIA FLARE federated learning with OpenDP differential privacy. Gradient updates cross boundaries; underlying data stays in jurisdiction.
Talk to Us
Our data sovereignty team can walk through how Sovereign maps to your specific compliance framework — GDPR, HIPAA, ITAR, or sovereign AI regulations.