For enterprises

Built for the work
that does not leave the building.

If your data falls under GDPR, HIPAA, GLP, FDA 21 CFR Part 11, MiFID II, BAIT, or any equivalent — and that data has to stay inside your network — Eldric is the AI platform that fits.

Where you draw the wall

Tenant guard, hash-chained audit, signed licenses.

Multi-tenant isolation

Every persisted artefact carries a tenant ID. The kernel returns 403 on any cross-tenant attempt. Superadmin escape hatch for legitimate cluster-internal calls only.

Hash-chained audit ledger

Every privileged operation appends an entry whose SHA-256 chain links to the previous entry. Verifiable offline.

Ed25519-signed licenses

The license file is signed and verifiable offline against an embedded public key. Worker validation falls back to the local file when the controller is unreachable.

Rolling cluster upgrade

Drain in-flight requests, snapshot state, install, verify SHA-256, restart, validate, move on. One alpha at a time, no downtime window.

OpenTelemetry export

Spans, counters, histograms via OTLP-HTTP. Low-cardinality path normalisation so dynamic IDs do not blow up your observability bill.

Federated learning

Multi-round federated training across worker nodes. Controller orchestrates; workers train on local data; gradients aggregate without sharing data.

Compliance, named

What we explicitly support today.

GDPR data stays on-prem HIPAA-aware GLP compliance mode FDA 21 CFR Part 11 e-records (Enterprise tier) Air-gap deployments Ed25519 signed releases

If your regulatory framework is named above, we have the matching toggle. If it is not, it does not mean we cannot support it — it means we have not had a customer ask yet. Write to office@eldric.ai.

Industrial real-time

Soft real-time, on the line.

Industrial AI often needs predictable latency, not just throughput. Eldric covers the soft-real-time and near-real-time tiers — the ones most production lines actually need. Hard real-time stays the PLC's job, and we say so.

Edge at the line

Deploy Eldric on a workstation or single server next to the PLC. The round-trip stays inside the building; no WAN hop, no cloud-call jitter.

Demo cluster: 41 ms p50 / 42 ms p99

Measured chat round-trip on the reference hardware. Latency stays flat as concurrency rises. Tag-to-action through the IoT worker sits in the same range.

xLSTM forecasting (port 8884)

The dedicated xLSTM daemon handles time-series forecast and policy workloads — designed for fast bounded-latency inference on industrial signals. Pre-trained weights still in distillation; runtime ships now.

Industrial protocols

OPC-UA, Modbus TCP/RTU, MQTT Sparkplug B in the IoT worker. Store-and-forward buffering through disconnect windows. Time-series historian for replay.

Native inferenced

Direct GGUF / xLSTM model loading via embedded llama.cpp — no Ollama hop, no extra process boundary. The shortest inference path Eldric ships.

What we don't claim

Hard real-time. Sub-millisecond deterministic control. RTOS-grade scheduling. Those are the PLC's job. Eldric sits above the PLC layer, where soft-real-time (10–100 ms) and near-real-time (event-driven streaming) are the right targets.

How procurement works.

We are a small Austrian company that does not run a self-service checkout. Every enterprise relationship starts with a conversation. The typical path:

  1. Discovery — you write to office@eldric.ai; we set up a call within two business days. We learn about your data, your regulator, your existing stack, and your constraints.
  2. Trial — you install a Standard-tier license in your staging environment. We provide a named contact who answers within one business day.
  3. Pilot — one team uses Eldric in production for thirty days. Real workload, real data, real users. We help debug.
  4. Production — Professional or Enterprise tier, multi-year, named support engineer for Enterprise.

We do not use third-party sales engineers. The person you talk to is the person who knows the codebase.

What you can ask for that nobody else gives you.

What we will not do.

Start a conversation See pricing