Building an AI-powered document pipeline platform that automates recurring compliance, fund, and regulatory reports — connecting to multiple data sources, validating the data, and generating professionally formatted Word and PDF documents, fully on-premise.
Python, AI agents, DSL-based template engine, Excel/SQL/REST/PDF connectors.
Deployed fully on-premise — no cloud LLM, no external API calls, no data leaving the client infrastructure. Fully GDPR-compliant by design.
5 seconds
Report generation time
100% on-prem
GDPR-compliant
4–8 weeks
Pipeline setup
IQAM Invest, an Austrian asset manager, produced semi-annual and annual fund reports entirely by hand. Analysts pulled data from multiple bank export formats, copied figures into Word templates, and cross-checked everything manually — a process that stretched across days and left no margin for error. One wrong number in a published fund report is one too many.
We developed PipeDok: an AI-powered document pipeline platform that connects directly to the data sources — Excel exports from multiple bank formats, SQL databases, REST APIs, and PDFs. The platform validates the incoming data with AI-assisted plausibility checks, then transfers the figures into professional Word templates using a DSL-based template engine. Every number in the finished report is traceable back to its exact source cell or database entry. Audit-proof by design.
What used to take IQAM Invest days now runs in seconds. The pipeline is live, the reports are consistent, and the team no longer touches a single cell manually. PipeDok is now in use across multiple report types — and new pipelines for other clients in regulated industries can be built and deployed in 4 to 8 weeks, not months.
Full traceability from output back to source — every figure linked to its exact cell or database record. Audit-proof reporting — no black boxes.
Excel, SQL databases, REST APIs, and PDFs — the pipeline consolidates any combination of structured, semi-structured, and unstructured data.
Sanity checks catch errors and outliers before they end up in the finished report. Subtotals verified, outliers flagged, missing values surfaced.
Layout changes without programming. One template update re-generates all reports automatically — no manual reformatting required.
No data leaves the organization. No cloud LLM. No US servers. Fully GDPR-compliant — built for regulated industries where data sovereignty is non-negotiable.
Workflow analysis in approximately one week. Custom pipeline built and deployed in 4–8 weeks using AI agents to analyze sources and build the automation.
Self-service Identity and Access Management for streamlined access control.
Migration of a mission-critical client access management platform from Credit Suisse to UBS Azure cloud infrastructure.
Greenfield rebuild of Austria's sovereign bond auction platform — managing EUR 315 billion in federal debt.
