Skip to main content

AI & Software · Software

AI Workspace Platform

Coming soonClosed early access

Multi-model assistant profiles, knowledge graph and federated search in one platform.

Request information Specifications
The Atlas knowledge-graph sphere with assistant profile panels and federated search rays.

Overview

The problem. In a mid-sized industrial or engineering organization, knowledge is scattered across three places: documents (specifications, drawings, manuals), chat and email history, and field data (SCADA, telemetry). Teams use different AI tools on personal accounts; data leaks out, costs stay invisible, and no one can audit which role accesses which model with what data. When the question is "did pump X have a similar fault last year, and what did we do?", the answer lives only in people's memory.

The solution. AI Workspace provides a single workspace inside the organization. A separate AI assistant profile is defined for each team and role, with independent model, permissions and context. The Atlas knowledge graph stores company data at the semantic level using a GraphRAG architecture; it extracts entities and relationships from documents and is queried in natural language. Federated search scans Atlas, documents and conversation history in a single query, with keyword, semantic and graph layers working together. A plugin architecture connects workflows (bid preparation, fault-report summarization, specification analysis) as modules. RBAC and usage/cost tracking give management visibility.

Knowledge joined with field data. Elmes's distinctive approach is to connect field telemetry to Atlas as well: with events from Telemetry & Process Intelligence, "fault history + maintenance document + live measurement" come together in a single query. This connection is on the development roadmap; the platform is currently in closed early access.

The Atlas knowledge-graph sphere, assistant profiles on frosted glass panels and light rays from a federated search bar.
Concept

This product is in development or field testing. Contact us for technical information, pilot use and a preliminary quotation.

Features

  • Multi-model assistant profiles: model, permissions and context are defined separately for each team and role.
  • Atlas knowledge graph — GraphRAG-based: extracts entities and relationships from documents; queried in natural language.
  • Federated search: Atlas, documents and conversation history in a single query across keyword, semantic and graph layers.
  • Plugin architecture: workflows such as bid preparation, fault-report summarization and specification analysis are added as modules.
  • RBAC + usage tracking: role-based access; usage and cost are tracked per role.
  • On-prem or cloud deployment: with an on-prem installation, data stays entirely inside the company.
  • TargetField-data bridge: events from Telemetry & Process Intelligence and data from the Wireless Gateway / RTU are written to Atlas semantically.
  • TargetModel mix: a local open-source model for confidential data and a proprietary API for general questions, governed by profile policy.
  • TargetAnswers with citations: every answer carries a reference to its source document or graph node.
  • TargetAudit log: all queries and tool calls are logged.

Specifications

Values marked “Target” are design targets, next-generation values or chip-vendor data; they are updated as measurement and certification are completed.

Deployment
Deployment modelOn-prem / cloud
Installation packageDocker / Kubernetes packageTarget
On-prem serverLinux; GPU for local modelsTarget
Cloud hostingData center in TürkiyeTarget
UpdatesVersioned container images with database migrationsTarget
ClientCurrent desktop and mobile browsersTarget
Server resourcesEstimate: core 4 vCPU / 16 GB RAM; GPU with ≥ 24 GB VRAM for a local LLMTarget
Architecture
Software stackTypeScript/Node services, PostgreSQL (pgvector) and a graph database; same PostgreSQL/MCP approach as Elmes DMTarget
Data modelTenant → workspace → profile; documents, chunks and embeddings; Atlas entities and relationships; conversations; audit; usage eventsTarget
Search tiersKeyword + semantic + graph
Integrations
ModelsOpen-source (local) + proprietary APIs
Provider abstractionCommon provider layer for switching between models (the pattern used in AutoPCB)Target
PluginsMCP tool serversTarget
Data sourcesFile upload in the first release; file system, SharePoint and email connectors; Telemetry & Process Intelligence APITarget
APIREST API and SDKTarget
Security & data
AuthenticationSSO (OIDC / LDAP)Target
AuthorizationRole-based access control (RBAC) + usage tracking
Data residency (on-prem)Data stays entirely on customer infrastructure
Data residency (cloud)In TürkiyeTarget
Proprietary model useCan be disabled per profileTarget
Audit logAll queries and tool calls are loggedTarget
Licensing & support
PricingOn quotation
License modelPer-user monthly subscription + on-prem installation feeTarget
Support (early access)Directly with the product team
Support (general availability)8×5, response within 1 business dayTarget
Roadmap
Stage 1Internal pilot on Elmes's own document archiveTarget
Stage 2Core platform, assistant profiles and document search (cloud)Target
Stage 3Atlas GraphRAG, federated search, RBAC and usage trackingTarget
Stage 4Plugins/MCP, Telemetry & Process Intelligence connector, on-prem packageTarget

Applications

  • R&D teams
  • Operations and maintenance
  • Field engineering
  • Management reporting
  • Bid and specification analysis
  • Technical document search

Scenarios targeted during early access:

  • R&D: semantic search across specifications, datasheets and past design documents; requirement generation with AutoPCB.
  • Operations and maintenance: fault history, maintenance manual and Telemetry & Process Intelligence anomaly records in a single query.
  • Field engineering: access to device documentation and commissioning notes via mobile or web.
  • Management: usage and cost reports, weekly summaries.
  • Bid and specification analysis: bringing Elmes's R&D consulting and specification analysis practice into the platform.

Compliance & documentation

AI Workspace is a software product and requires no mandatory product certification. The data-protection and security documents needed for enterprise use are being prepared according to the plan below.

  • KVKK privacy notice and data processing agreement
    Required for cloud use
    Target
  • Security policy and penetration test report
    Target
  • ISO 27001
    Long term, depending on enterprise customer demand
    Target
  • Third-party model license inventory
    Target
  • User and administrator guide
    Target
  • Mandatory product certification
    Software product; no mandatory certificate
    Out of scope

Frequently asked questions

When will the product be available?

The platform is in closed early access; a general availability date has not been set yet. Pilot participation is offered on quotation.

Does our data leave the company?

Not with an on-prem installation; the platform runs entirely in-house with local open-source models. For cloud installations, the option to disable proprietary models per profile is planned.

Which models are supported?

Both open-source and proprietary models are supported. A concrete list of supported models has not been published yet.

What does the Atlas knowledge graph solve?

It extracts the relationships between documents (product ↔ fault ↔ customer ↔ solution) and makes them queryable, revealing links that classic keyword search cannot find.

Can we connect our SCADA or telemetry data?

This feature is on the development roadmap and is being designed together with Telemetry & Process Intelligence.

Does it integrate with our document system (SharePoint, file server)?

The first release works with file upload; SharePoint, file system and email connectors are planned.

How is it priced?

Pricing is on quotation. A model combining a per-user monthly subscription with an on-prem installation fee is planned.

How is it different from large general-purpose AI platforms?

It is aimed at SMEs and mid-sized industrial companies. The difference lies less in general knowledge access and more in integration with field data (telemetry, fault history, maintenance documents).

Get information about this product

Our engineering team replies within 24 hours.

Request information +90 (312) 395 85 86

Related products