AI-Driven Public Services: Automating Document Processing and Citizen Support
Scheduled sessions
Modernizing Public IT Infrastructure: Transition from legacy, manual processes to AI-driven workflows while strictly maintaining data sovereignty and citizen privacy.
On-Premise AI Deployment: Learn how to run state-of-the-art open-source Large Language Models (like Llama 3 or Mistral) entirely within your institution's secure network, ensuring zero data leakage.
Document Automation: Combine Optical Character Recognition (OCR) with Vision-Language Models to automatically extract, categorize, and validate information from citizen applications and official forms.
Internal Virtual Assistants: Build Retrieval-Augmented Generation (RAG) systems trained on official gazettes, legal frameworks, and internal procedures to assist public servants in finding accurate answers instantly.
Who it’s for: Public Sector IT Managers, System Architects, and Technical Business Analysts responsible for modernizing government tech stacks securely.
Skills You Will Learn
Curriculum
Data Sovereignty and On-Premise LLMs
- The imperative for local AI: Why public institutions must avoid external APIs for PII (Personally Identifiable Information)
- Evaluating open-source models (Llama 3, Mistral) for government use cases
- Deploying LLMs locally: Utilizing Ollama and vLLM for high-throughput, secure inference
- Lab: Spinning up an air-gapped LLM environment and testing inference speeds
Automating Bureaucracy: Document Processing
- Modern OCR pipelines: Extracting text from scanned citizen requests and handwritten forms
- Information Extraction: Using LLMs to pull structured data (JSON) from unstructured legal or administrative documents
- Categorization and Triage: Automatically routing citizen requests to the correct department
- Lab: Building an automated pipeline that extracts PII from sample ID cards and application forms securely
Building Secure Internal Assistants (RAG)
- RAG architecture for public administration: Querying the Official Gazette or internal SOPs
- Setting up local vector databases (e.g., PostgreSQL with pgvector) for document embedding
- Ensuring factual accuracy: Techniques to strictly ground the LLM in retrieved documents
- Lab: Creating an internal chat assistant that answers procedural questions based exclusively on uploaded policy PDFs
System Integration and Workflow Automation
- Connecting AI components to existing legacy public sector databases (SQL integration)
- Event-driven architecture: Triggering AI evaluation upon form submission via citizen portals
- Human-in-the-Loop (HITL) design: Ensuring civil servants review AI-extracted data before final approval
- Lab: End-to-end integration mapping an automated document review workflow
Course Day Structure
- Part 1: 09:00–10:30
- Break: 10:30–10:45
- Part 2: 10:45–12:15
- Lunch break: 12:15–13:15
- Part 3: 13:15–15:15
- Break: 15:15–15:30
- Part 4: 15:30–17:30