Back to projects
AI AgentsFeature Extraction

Danish Public Administration: AI Business Automation

A public-sector document automation workflow using microservices and AI agents under strict process constraints.

Role
AI automation and backend engineering
Focus
AI Agents / Feature Extraction
Stack
Python, FastAPI, Kubernetes

Context

The shape of the problem

Public administration workflows often depend on semi-structured documents, external lookups, and explicit business rules. The challenge was to automate targeted parts of that work without losing traceability or quality control.

Build

What I actually built

I contributed to a microservices architecture with FastAPI, Kubernetes, PostgreSQL, and specialized extraction agents. The system combined document processing, scraping, and rule-driven case handling so the automation could be measured against the real process.

Engineering

Key decisions

01

Handling complex, semi-structured government documents

Developed specialized document processing microservices with FastAPI

02

Meeting strict security and compliance requirements

Implemented a secure, containerized architecture with Kubernetes

03

Ensuring high accuracy in automated extraction

Created custom extraction agents following process mined business rules

Outcome

What changed

  1. 01Achieved 96% automation rate for targeted workflows
  2. 02Clearly defined business logic for case handling and document processing
  3. 03Quality assurance agent on existing cases.

Stack

Tools and systems

PythonFastAPIKubernetesPostgreSQLDockerDevOpsAzure Pipelines

I build delightful systems.

© 2026 Kristian Ernst.