Everything you need to know about Mergic

Mergic is a GPU-accelerated multimodal AI platform that unifies imaging, genomics, and clinical data into a single cross-modal reasoning layer. It ingests DICOM, FASTQ, VCF, EHR data, and other biomedical formats, runs GPU-native inference pipelines using NVIDIA Clara, MONAI, Parabricks, and BioNeMo, and generates agentic clinical summaries with evidence citations and human signoff workflows.

Mergic is built around Clara (medical imaging workflows), MONAI (imaging AI), Holoscan (real-time streaming), Parabricks (genomics pipelines), BioNeMo (molecular generation and protein folding), NeMo (LLM reasoning), and NIM (inference microservices). These are integrated natively, not as aftermarket add-ons.

Yes. Many teams begin with imaging-only or genomics-only workflows and gradually expand into cross-modal pipelines. The shared data flywheel (Mergic Atlas) makes it straightforward to add new modalities without re-ingesting existing datasets.

For training workloads, we recommend NVIDIA A100, H100, or DGX systems. Inference can run on T4, L4, A10G, and above. The platform automatically detects available hardware and optimizes kernel selection for throughput and latency.

Yes. Enterprise customers can fine-tune pre-trained foundation models on their own datasets or bring custom models via the Mergic model registry. Citadel governance tracks model lineage, versioning, and promotion workflows from development to production.

Mergic encrypts data at rest (AES-256), in transit (TLS 1.3), and in compute (confidential computing enclaves where available). Private deployments keep all patient data inside your VPC or on-prem perimeter. De-identification pipelines automatically strip PHI from imaging metadata, genomic headers, and clinical text. Audit trails log every data access with cryptographic integrity verification.

Yes. Mergic maintains SOC 2 Type II certification and supports HIPAA BAA execution for enterprise customers. Deployments include built-in controls for encryption, access logging, de-identification, and role-based access that satisfy HIPAA Security Rule requirements.

Beyond HIPAA and SOC 2, Mergic provides controls aligned with GDPR (data portability, right to erasure), HITRUST (healthcare risk management), FDA 21 CFR Part 11 (electronic records and signatures), and ISO 27001. Annual third-party penetration testing and continuous compliance monitoring are included.

Yes. Every inference, data access, human signoff, and policy change is logged to an immutable append-only ledger. Audit records include model version, input data provenance, confidence scores, and cryptographic signatures for tamper detection. Records are retained for 7 years by default.

In private deployments (VPC or on-prem), only your authorized users have access. Mergic engineering does not access customer PHI except during explicitly scoped support engagements with signed data-access agreements. Multi-tenant cloud deployments use tenant-level isolation with customer-managed encryption keys.

Mergic supports cloud (AWS, Azure, GCP), dedicated VPC (single-tenant with customer-managed keys), on-premises DGX (full stack in your data center), and hybrid (split inference and training across on-prem and cloud). Air-gapped installations are supported for on-prem deployments with no external network dependencies.

Yes. The on-prem DGX deployment supports fully air-gapped installations with offline model updates via USB or secure file transfer, local container registries, and no external network dependencies for inference or governance operations.

Cloud and VPC deployments typically complete in 1-2 weeks including integration with PACS/LIS/EHR systems. On-prem DGX deployments take 4-6 weeks including hardware provisioning, installation, and validation. Hybrid architectures add 2-4 weeks for secure data synchronization setup.

Enterprise contracts include white-glove deployment assistance: architecture planning, integration development, performance tuning, security hardening, and training for IT and clinical teams. Ongoing support includes dedicated Slack channels, 24/7 incident response, and quarterly business reviews.

Cloud and VPC deployments include 99.9% uptime SLA with proactive monitoring, automated failover, and 24/7 incident response. On-prem deployments include installation validation and support SLA but uptime depends on customer infrastructure and operations.

Pricing depends on deployment type, modalities used, compute requirements, and data volume. Cloud deployments use usage-based pricing (inference calls, storage, GPU hours). VPC and on-prem deployments use annual licensing with tiered capacity bands. Contact sales for a custom quote tailored to your workflows.

Yes. We offer 30-90 day pilots for qualified healthcare organizations and research institutions. Pilots include limited-scope deployment, sample data processing, and success metrics definition. Pilot credits can roll into production contracts upon conversion.

Enterprise contracts include deployment assistance, dedicated support channels, SLA commitments, security and compliance documentation, model fine-tuning credits, integration development hours, and quarterly business reviews. Volume discounts and multi-year terms are available.

Yes. Academic medical centers, university research labs, and qualified non-profit healthcare organizations are eligible for discounted pricing. Contact sales with your institution details for a custom academic quote.

Mergic is available via AWS Marketplace, Azure Marketplace, and select healthcare technology resellers. Marketplace purchases support simplified procurement and consolidated billing for organizations already using cloud providers.

Mergic provides a platform and tools that support clinical-grade AI workflows, but it is not marketed as a cleared medical device. Customers deploying Mergic for diagnostic or treatment decision use cases are responsible for validation, regulatory pathway determination, and obtaining necessary clearances based on their intended use.

All Mergic foundation models are trained on peer-reviewed benchmarks and validated against published datasets. Customers deploying for clinical decision support should conduct internal validation studies on representative patient populations and establish performance thresholds aligned with their risk management frameworks.

Yes. All AI-generated findings include mandatory human signoff checkpoints before recommendations reach patient records. Configurable escalation rules route uncertain or high-acuity cases to specialist review, and all signoffs are logged in immutable audit trails.

Every inference includes calibrated confidence scores and provenance links to input data. When confidence falls below configurable thresholds, cases automatically escalate to human review. Models are trained with uncertainty quantification and regularly re-calibrated against ground-truth clinical outcomes.

Yes. Mergic connects bidirectionally with PACS, LIS, and EHR systems using standard protocols (DICOMweb, FHIR R4, HL7v2). Structured reports, annotations, and AI findings can be pushed back into clinical systems and surfaced in existing radiologist, genomics, and clinician review queues.

Mergic natively ingests DICOM, WSI (SVS, NDPI, MRXS), FASTQ, BAM, CRAM, VCF, HL7v2, FHIR R4, PDF, and structured CSV/Parquet. Custom parsers can be added via the plugin SDK for proprietary instrument formats or research-specific file types.

Mergic enforces consent policies at ingestion time, automatically tagging data by governance rules. Consent metadata travels with data through all pipelines, and policy engines block processing for patients who have not granted appropriate consent flags (e.g., research use, AI analysis).

Yes. Mergic supports both clinical production workflows and research use cases. Research teams can ingest de-identified cohorts, run experimental models, and track lineage separately from production pipelines. Citadel governance maintains separate policy boundaries between clinical and research environments.

Retention policies are customer-configurable. By default, raw ingested data is retained for 90 days, processed outputs and embeddings for 1 year, and audit logs for 7 years. Custom retention schedules can be defined per data type, modality, or patient cohort.

Yes. Mergic supports full data export in original formats via API or bulk download. Data deletion requests (e.g., for GDPR right to erasure or patient withdrawal) trigger automated workflows that purge data from storage, embeddings, and backups, with cryptographic deletion verification.

Default rate limits are 10,000 requests per minute for inference endpoints and 50,000 requests per minute for read-only queries. Enterprise customers can negotiate higher limits based on workload requirements. Rate limit headers are included in all API responses for client-side throttling.