CLIQUE
Problem Product How It Works ROI Contact Control Plane Studio See both demos Request Demo
On-Premise AI Orchestration

The Operating System for Hospital Compute

Clique is software that installs in the hospital data center and schedules AI workloads across the CPUs, GPUs, and storage already sitting there. IT gets a live control plane; clinicians get a no-code model trainer. Patient data never leaves the building.

90% of hospital servers sit idle
$2.4M avg. annual cloud spend per AMC
0 GB patient data leaves the hospital
Scroll

The same engine,
a UI shaped to the job.

For IT, CIOs, compliance

Control Plane

A dense, real-time view of every GPU, workload, vendor container, enclave, and audit event in your hospital — with role-scoped views for Clinician, IT, and Researcher.

  • Live telemetry across every node, partition, and workload — one pane of glass.
  • Continuous HIPAA and SOC 2 checks with one-click PDF export for auditors.
  • Emergency priority lane that preempts batch jobs for outbreaks and clinical urgency.
Explore Control Plane
For clinicians, researchers

Studio

Drag in a labeled dataset, pick the task, and Clique auto-configures the pipeline and trains the model in a self-spinning container. No code. No config files. No DevOps.

  • Drop a labeled dataset, pick a task type — segmentation, classification, detection.
  • Auto-configured pipeline with nnU-Net-inspired heuristics; live training you can watch.
  • Versioned artifact handed off to IT for review, deployment, and audit trail.
Explore Studio

Hospitals pay twice: idle servers
at home, rented GPUs in the cloud

~10%

Server Utilization

Hospital servers run at a fraction of capacity. IT provisions for peak load, and the rest depreciates as pure cost.

2-3x

Cloud Markup

Egress fees, compute markup, and compliance overhead make cloud AI economically punishing for healthcare.

Siloed

Fragmented Compute

Departments hoard GPUs with no safe way to share, schedule, or audit workloads across the organization.

Hospitals don't need more hardware or another cloud contract. They need software that finds every idle GPU in the building, schedules AI jobs onto it by clinical priority, and proves to compliance that no patient data ever left.

One platform. Two sides.
Zero trust required.

Clique OS
Discovery Auto-inventory all CPUs, GPUs, storage
Scheduling Clinical-priority-aware orchestration
GPU Partition NVIDIA MIG for performance isolation — one GPU, many workloads
DICOM Bridge Native PACS integration for clinical AI
Remote Attestation Cryptographic proof of enclave integrity before data enters
Confidential Compute Intel TDX + AMD SEV-SNP + NVIDIA H100 CC
Dashboard Utilization, savings, compliance status

See everything. Control everything.

Clique Control Plane
Overview
Workloads
Resources
Compliance
Settings
GPU Utilization 87%
Active Jobs 24
Savings this month $113K
WorkloadPriorityGPUStatus
Stroke Detection v3CriticalA100 #2Running
Path. Slide AnalysisHighA100 #1Running
Research — Cohort 7BatchV100 #3Queued
NLP Discharge NotesHighA100 #1 (MIG)Running
Total Workloads 31
Queued 7
Avg. Latency 48ms
WorkloadDepartmentRuntimeStatus
Stroke Detection v3Radiology2h 14mRunning
Path. Slide AnalysisPathology6h 02mRunning
Chest X-Ray TriageEmergency0h 47mRunning
Genomics PipelineResearchQueued
GPUs Online 6 / 6
CPU Utilization 62%
Storage Used 184TB
DeviceTypeMIG SlicesHealth
A100 #1NVIDIA A100 80GB3g.20gb × 2Healthy
A100 #2NVIDIA A100 80GBFull deviceHealthy
V100 #1NVIDIA V100 32GBN/AHealthy
V100 #2NVIDIA V100 32GBN/AStandby
HIPAA Status Compliant
Enclaves Attested 6 / 6
Audit Events 12,847
CheckFrameworkLast RunResult
Data residencyHIPAA2 min agoPass
Encryption at restSOC 25 min agoPass
Enclave attestationTEE1 min agoPass
Access control auditHIPAA12 min agoPass
Scheduling Policy Clinical > Research > Batch
GPU Partitioning MIG enabled on A100 #1
DICOM Endpoint pacs.hospital.internal:11112
Attestation Interval Every 60 seconds
Audit Log Retention 7 years (HIPAA)

Hospitals don't need more hardware. They don't need another cloud contract. They need an operating system for the infrastructure they already own.

From 10% server utilization
to 70%+ in weeks

01

Deploy Agent

Lightweight agent installs on existing servers. Discovers and inventories all heterogeneous compute — CPUs, GPUs, storage — in hours, not months.

02

Connect Workflows

DICOM bridge integrates with your PACS. Radiologists and pathologists see AI results in their existing tools. Zero workflow disruption.

03

Configure & Secure

Define TEE boundaries per workload. Hardware attestation verifies enclave integrity. AI vendors deploy models into sealed environments.

04

Orchestrate

Clinical-priority scheduling goes live. Stroke detection preempts research batches. GPU partitioning serves multiple workloads. Utilization goes from 10% to 70%+.

clique-cli
$ clique init --discover
Scanning network for compute resources...
  Found: 4x NVIDIA A100 · 2x V100 · 128 CPU cores
  Storage: 48TB NVMe · 220TB SAN
Hardware inventory complete
$ clique schedule --priority clinical
Priority scheduler activated
  Policy: clinical > research > batch
  MIG: A100#1 partitioned → 3g.20gb × 2
$ clique status
  GPU util: 87% ↑ from 10%
  Jobs: 24 active · 3 queued
  Savings: $113,200 /month
All enclaves attested · HIPAA compliant

The math is simple

ROI Model: 500-Bed Hospital

Current Cloud Cost (annual)

GPU inference (800K studies × $1.35) $1.08M
Egress fees (1.4 PB × $0.09/GB) $126K
Training runs (custom models, 5/yr) $175K
Managed services, MLOps, compliance $340K
Total cloud spend $1.72M

With Clique (annual)

Clique platform license (500 beds) $250K
Incremental GPU amortized (5yr) $60K
Power, cooling, rack space $50K
Cloud burst (peak overflow only) $0
Total on-prem spend $360K
~$1.36M Net savings / year
·
~2 mo Payback period
·
79% Cost reduction vs. cloud

Built for the demands of

HIPAA GDPR SOC 2 Type II DICOM Native NVIDIA MIG TEE Supported
Initializing fabric
Breadth-first discovery across every node.
Then every pair, every path — one clique.

Your servers are running at 10%.
Let's fix that.

Currently onboarding design partners in the Bay Area and Northeast