Grae SI

Stack

Ten layers. Three kinds of ownership.

Grae SI is the studio for the model, the evaluation packet, and the offline runtime. It is not the EHR, the safety database, the quality system, or the committee that signs. The layers below are the default. Each sandbox profile says where that default moves, and how many person-weeks the buyer should still plan after the runtime is up.

  • Grae SI providesGrae SI provides the layer
  • SharedShared. Grae SI supports part of it.
  • You ownYou own the layer after install

  1. Layer 1

    Infrastructure and compute

    You own

    A runtime bundle and a sizing note for one GPU host or a small private pool. The machine, the network zone, power, and the vendor who racks it.

    Buyer effort after the runtime is up: 1–3 person-weeks

  2. Layer 2

    Data ingestion and prep

    Shared

    A checklist of what to collect, a sample upload, and a readiness score. The pipeline from the system of record into an approved extract.

    Buyer effort after the runtime is up: 2–5 person-weeks

  3. Layer 3

    Data controls and governance

    Shared

    Masking preview, data-class prompts, and a documentation list tied to the tier. Access control, residency, retention, and the policy that makes those real.

    Buyer effort after the runtime is up: 1–3 person-weeks

  4. Layer 4

    Model selection and training

    Grae SI provides

    Catalog, license flags, one job spec, and adapters for local, managed, or your cluster. The choice of base model, the license acceptance, and the gold data you are willing to use.

    Buyer effort after the runtime is up: 0.5–1.5 person-weeks

  5. Layer 5

    Evaluation and validation

    Shared

    A tier-matched suite, thresholds, and an exportable report. The labeled set, the acceptance call, and any statistics your quality system requires on top.

    Buyer effort after the runtime is up: 1–3 person-weeks

  6. Layer 6

    Agents and orchestration

    Shared

    Compliance agents that classify the use case and name the frameworks to prepare for. Any production agent that calls tools inside your applications.

    Buyer effort after the runtime is up: 1–2 person-weeks

  7. Layer 7

    Deployment and serving

    Grae SI provides

    Safetensors, GGUF, a model card, and an offline bundle for Ollama, vLLM, or llama.cpp. Moving that bundle into the zone and keeping the host patched.

    Buyer effort after the runtime is up: 0.5–1.5 person-weeks

  8. Layer 8

    Monitoring and drift

    Shared

    Registry lineage: version, dataset, eval result, and the package that was released. Live traffic checks, drift alarms, and the decision to pull a version.

    Buyer effort after the runtime is up: 1–3 person-weeks

  9. Layer 9

    Application and integration

    You own

    Nothing inside your system of record. The model is a service you call. The screen, the workflow, and the audit log in the tool people already use.

    Buyer effort after the runtime is up: 2–8 person-weeks

  10. Layer 10

    Organizational validation and sign-off

    You own

    A packet a reviewer can read. Not a certification, and not a GxP release. The quality owner, the human review, and the record in your quality system.

    Buyer effort after the runtime is up: 1–4 person-weeks

What “provides” means

Grae SI provides
The catalog and license flags, one training job spec, the tier-matched evaluation report, the model card, and the runtime bundle. You still choose the base model and accept its license.
Shared
Data checklists, masking previews, and compliance agents. Your pipelines, access reviews, and production monitors do the rest. A shared layer is unfinished if only one side shows up.
You own
The host, the system of record, and the signature in your quality system. Person-weeks on each profile are this work, after deploy, not a claim that Grae SI installs your EHR.

Default layer count: 10. Profiles reuse these ids so a buyer can compare a clinic with a twin program without learning a second diagram.

Open a profile