For research institutions

Public research.
Public infrastructure.
Public outcomes.

Research labs, public-research HPC fabrics, AI factories, member-state sovereign-compute programmes — the institutions running AI workloads on hardware they own. Eldric is the operating system that fits the way you actually work: on-prem, multi-tenant, audited, federated across sites, with 140+ scientific APIs out of the box.

Sovereign compute by default

Your hardware. Your data. Your jurisdiction.

Hardware-agnostic

The same operating system on HPE, Eviden BullSequana, Dell, NVIDIA, mixed clusters. No vendor lock-in at the AI layer. We sell the platform above the iron, not the iron itself.

On-prem first

The deployment lives in your facility, on your network, behind your firewall. No mandatory cloud calls. No data egress for inference, training, or RAG.

Multi-site federation

Same kernel from a single rack to a cross-site cluster. Federation across data centres, with raft-based controller consensus and asynchronous cross-site replication.

Standards-grade audit

The receipts your auditor will ask for.

Audit ledger

Every privileged operation appends an entry to a tamper-evident, cryptographically-chained ledger. Verifiable offline.

GLP and 21 CFR Part 11 modes

Pre-clinical and regulated-research toggles. Electronic-records / electronic-signatures support, sample-tracking with traceability, validated state transitions.

Signed releases

Every release is cryptographically signed. The repository is signed. The license file is signed and verifiable offline. Reproducible install path for air-gap deployments.

If your regulator is named above, we ship the matching toggle today. If not, write to office@eldric.ai — it usually means we have not had the conversation yet, not that we cannot build it.

Multi-tenant by design

Shared infrastructure. Cleanly separated tenants.

Per-team isolation

Every persisted artefact — vector embedding, agent session, knowledge base entry, training job — carries a tenant identifier. The kernel returns a hard 403 on any cross-tenant attempt. Useful when a single cluster serves multiple labs, departments, or consortium partners.

Role-based access

Viewer, developer, admin, superadmin per tenant; cross-tenant administration only via the platform superadmin role. Audit trail records who acted on whose behalf.

Quotas and tiers

Per-tenant quotas for sessions, knowledge bases, training rounds, storage. License tiers cleanly enforced; no honour-system caps.

Per-tenant theming

White-label the chat shell per tenant. Useful when separate departments or member-state programmes share infrastructure but want distinct identities.

Scientific tooling

140+ scientific APIs out of the box.

The science worker carries dispatch tables for the data sources your researchers already cite, plus the analysis tooling they already use. Custom sources plug in through a registry-driven extension path; no patching the core.

Astronomy, physics, earth

NASA, ESA, JAXA, ISRO, Roscosmos. CERN, DESY, Fermilab. LIGO, GWOSC, ESO. USGS earthquakes and volcanoes, NOAA weather, ice cores, glaciers.

Bioinformatics & pharma

Ensembl, ENCODE, GTEx, HMDB for genomics. BLAST, variant calling, sequence alignment. CRISPR guide-RNA design, base / prime editing, off-target analysis. Molecular docking, ADMET, AlphaFold integration.

Medical, climate, archaeology

Clinical trials, WHO, GWAS, OpenFDA, PubMed. GBIF, OBIS, air quality. PaleoBio and Open Context for archaeology. Materials Project, COD, NIST chemistry.

LIMS, audit trails

Sample tracking, experiment management, reagent inventory, GLP-aware state transitions. Audit-ready provenance for every step.

Federated training

Multi-round federated training across worker nodes. Controllers orchestrate; workers train on local data; gradients aggregate. Useful when consortium partners cannot pool data but can share model improvements.

Custom data sources

Add an institution-specific data source through the registry-driven plugin path. No core patch needed; the LLM dispatches against it on the next call.

Scale

From a workstation to a multi-site fabric.

Eldric runs on a Raspberry Pi for a single research-group demo, on a workstation for a department, on a single-site cluster for an institution, and on a federated multi-site fabric for an AI factory or member-state programme. The same operating system at every scale; the topology you actually need adjusts via the admin GUI, not a different product.

Single workstation

One install, one host. Suitable for a research group exploring on-prem AI for the first time, or for a sensitive workload that should never share hardware.

Single-site cluster

Controller plus a handful of inference nodes plus data workers. Suitable for an institute or department running shared AI services for ~hundreds of researchers.

Cross-site federation

Multi-controller raft cluster forming across geographic sites. Suitable for a national HPC fabric, a consortium of research institutions, or an AI factory spanning multiple data centres.

Air-gap and partial-air-gap

Signed RPM repository on a USB key or behind an internal mirror. Installs and runs without internet access. Useful for facilities where outbound connectivity is policy-prohibited.

How procurement works.

Eldric is a small Austrian company. Every research-institution relationship starts with a conversation, not a self-service checkout. The typical path:

  1. Discovery — you write to office@eldric.ai; we set up a call within two business days to learn about your facility, your existing HPC stack, your regulator, and the workloads you want to bring on-prem.
  2. Trial — Standard-tier license in your staging environment. Named contact for one-business-day response.
  3. Pilot — one group, thirty days, real workload, real data. We help debug.
  4. Production — Professional or Enterprise tier, multi-year. Source-code access available at Enterprise / Custom tier; on-site deployment assistance available; air-gap install supported.

The person you talk to is the person who knows the codebase. No third-party sales engineers.

Talk to us See pricing