Technology

How ASF Core is built

One governed platform: shared engineering context, AI grounded in your sources, one review model for every output and deployment that fits regulated teams.

Architecture

Four layers, one source of truth

  1. 01

    Context

    Project, item, system and device definitions, requirements and the knowledge base are captured once and reused by every analysis.

  2. 02

    Grounded AI

    Hybrid retrieval over your documents and standards using document tree, vector and keyword search, with documents selectable per workspace. The Explain panel shows each AI row’s rationale, confidence and source passages with page and relevance score (and the scoring method) where the generation recorded them, and says so when it did not; deterministic checks catch inconsistent ratings, and missing inputs are shown as not computable rather than filled in.

  3. 03

    Review and governance

    Analysis rows, goals, requirements, claims and controls share one approval model (AI suggested, in review, approved) with roles, review queues and an audit trail. Regeneration never overwrites approved work; baselines and a side-by-side diff show added and removed rows, changed tree nodes and changed requirement wording.

  4. 04

    Traceability

    Upstream and downstream links between hazards, goals, requirements and evidence, with change impact visible before approval.

In the product

Grounding and review, in the product

AI reasoning you can inspect: Every AI-suggested row explains its rationale, inputs and confidence before an engineer approves it.

Design principles

Built for work an assessor will read

Engineers decide

AI drafts; nothing becomes part of a work product without a named reviewer’s approval.

Sources over fluency

Suggestions are judged by the passages they cite, not by how convincing they sound.

Never guessed

Missing inputs show as “not computable” or “not recorded”. Probabilities, PMHF and ratings are never defaulted.

Industry-aware context

Industry profiles bring their own terminology, standards mapping and artifact sets on one shared core.

One model for every output

Every artifact is a registry entry per industry profile and standard. Tables, records, diagrams and fault/event trees open in shared workspaces with the same status model and export path.

Open exchange

ReqIF 1.0.1, requirements connectors for Polarion, Codebeamer, DOORS Next and Jama, Jira work items and medini analyze exchange (early access).

No certification claims

ASF supports the engineering work. Certification and conformity decisions stay with qualified assessors.

Deployment and data

Run it where your data has to stay

Managed cloud

The fastest way to start a pilot. EU hosting is agreed during deployment planning.

Private cloud / VPC

Dedicated deployment in your cloud account with data residency under your control.

Self-hosted

On-premises installation including air-gapped options and bring-your-own-model support.

Deployment, model choice and support terms are agreed per plan.

Bring your current workflow. We’ll show you how ASF can improve it.

45 minutes with the team, using an item or product of your choice. We reply within one business day.