PLATFORM
The technical foundation for agents that work with documents.
feld.ai connects documents, reference data, case files and business systems. We use this foundation to build agents for invoice intake, digital mailrooms and case-file work. Classification and extraction are building blocks; the result in the existing workflow is what matters.
Deployment and data flow
You choose where processing runs: on our infrastructure in Feldkirch or in your own environment. For each deployment, we agree the data flow, access rules and retention with you.
Own hardware in Austria
Our GPU servers are located in Feldkirch, Vorarlberg. No AWS, no Azure, no GCP. Physical control over every rack.
Zero data flow to big tech
No API calls to OpenAI, Google, or Amazon. All models run locally. Your data is never used to train external AI.
Managed or on-premises
feld.ai cloud, deployment in your infrastructure, or full on-premises installation. You decide where your data lives.
Data protection by deployment design
Data flows, access, retention and model use are defined for each project. The service can run in Austria or in your environment.
The formats your workflow actually receives
feld.ai processes PDFs, scans, emails, images and Office files, among other formats. We verify which formats can be automated reliably against the actual document set.
PDFs, scans and e-invoices
Digital PDFs, scans, ZUGFeRD, XRechnung and Swiss QR-bills in one reliable intake workflow.
Excel & spreadsheets
Structured and unstructured worksheets, multiple tabs, nested headers. No format assumptions.
Word & PowerPoint
Text documents, reports, presentations. Layout-aware extraction that understands structure.
Images & photos
Document photos, screenshots, whiteboard captures. Multimodal AI recognizes content even at low quality.
Emails & attachments
MSG, EML, and arbitrarily nested attachments. All incoming mail processed in a single step.
Multilingual
German, French, Italian, English, and more. Language detection is automatic, no manual tagging required.
Building blocks for the complete workflow
The platform does more than read content. It assigns documents to cases, matches them against existing data, keeps exceptions traceable and delivers approved results to the system of record.
Classification
Automatically identify document types: invoice, contract, inquiry, notice. Configurable for your document types and categories.
Extraction
Define fields, AI extracts. From invoice amounts to clauses in 200-page contracts.
Filtered worklists
Filter by system fields and extracted values, save views and export the current selection to Excel.
Source references
Extracted values link back to their location in the source document.
Master data matching
Automatic matching with suppliers, IBANs, accounts, cost centres. Structured data, ready to use.
Case files and matters
Bring documents into case files and matters, keep required data visible, and answer questions with source references.
Semantic search & chat
Ask questions about your document archives in natural language. Answers with source references.
Completeness and review
Missing required data becomes a clear worklist. Values can be reviewed and corrected directly against the source document.
Redaction
Identify sensitive data and custom search terms, review each match and export a redacted PDF.
From input to delivery
Integration
The public REST API covers reusable document and invoice capabilities. Complete case-file and agent solutions are configured for the workflow and connected through the agreed integration.
Public API
Project-scoped keys, uploads and asynchronous jobs. API documentation
Document jobs
Upload, split, classify, extract and summarise documents.
Invoices and reference data
Ingest electronic invoices, produce invoice data and optionally match it against versioned master and reference data.
Case-file and agent solutions
Intake, filing, required data, checklists, notes, chat, exports and document generation as one configured workflow.
Run without a US cloud
In the feld.ai-managed setup, the application and AI models run on our infrastructure in Austria. Alternatively, we deploy the service in your environment.
Data flows, access, retention and model use are defined for each project. No external model API is required.
- Processing on feld.ai infrastructure in Austria or in your environment
- No external model API required
- Data flows, access and retention defined for each project
- Connection through the integration agreed for the workflow
Model architecture: three tiers, all local
feld.ai runs all AI models on its own GPU servers in Feldkirch. Zero API calls to OpenAI, Anthropic, Google Vertex, or Azure OpenAI. Three tiers:
Large open-weights models
For document understanding, chat, and reasoning. Model families under evaluation: DeepSeek, GLM, Qwen, Llama, Mistral, Kimi, Gemma, Phi. Downloaded and run locally.
Proprietary small models
Built from the ground up by the feld.ai team. Graph-based models for Excel parsing: table detection, layout understanding, cell classification, mapping to ontologies.
Fine-tuned specialist models
Fine-tuned and re-implemented models for specific document types. Optimized for high precision and throughput in production.
All three tiers run air-gapped on own servers. No model has internet access. No model sends telemetry. Full disclosure: Sovereignty.