CompliScan-AI

EASA Part-145 MOE Automated Compliance Analysis System

Mission

Automate EASA Part-145 MOE compliance verification for Traficom aviation inspectors, reducing manual audit time from days to minutes.

---

Key Features

1. AI-Powered Analysis

• Semantic Matching: Compares MOE content against all EASA Part-145 clauses
• Intelligent Judgment: Google Gemini AI evaluates complex compliance cases
• Evidence Mapping: Links specific MOE paragraphs to each requirement

2. Structured Reporting

Each requirement analyzed with comprehensive JSON response data:

API Response Structure:

{
  "compliance_status": "non",        // full | partial | non
  "requirement_id": "AMC_145_A_10",
  "finding_level": "Level 2",        // Level 1 | Level 2 | Observation | Recommendation
  "justification": "The provided MOE excerpts do not contain the definition of 'Line maintenance'",
  "missing_elements": [
    "Definition of 'Line maintenance'",
    "Specific activities included in line maintenance"
  ],
  "recommended_actions": [
    "Update MOE to include the definition of 'Line maintenance' and its scope as per AMC 145.A.10"
  ],
  "evidence": [
    {
      "moe_paragraph_id": 692,
      "relevant_excerpt": "This section specifies the categories and requirements...",
      "similarity_score": 0.87,
      "rerank_score": 0.87
    }
  ]
}

UI Display (Dropdown Format):

1. Compliance Status - Visual badge (Full/Partial/Non)
2. Requirement - ID & title (e.g., AMC_145_A_10: Line Maintenance)
3. Justification - AI reasoning for the judgment
4. Missing Elements - List of gaps found
5. Recommended Actions - Specific remediation steps
6. Evidence - MOE paragraphs with similarity scores (expandable)

3. Inspector-Optimized UI

• Dropdown Design: Clean overview with expandable details
• Priority Ordering: Critical findings first
• Finding Levels: Level 1/2, Observations, Recommendations
• Pagination: 10 requirements per page for easy review

---

Architecture

Tech Stack

• Backend: Java 21, Spring Boot 3.5.7
• Frontend: Next.js 16, React 19
• AI: Google Gemini (embeddings + judgment)
• Database: PostgreSQL 17

---

Project Structure

compliscan-ai/
├── backend/          # Spring Boot API
├── frontend/         # Next.js UI
├── uploads/moe/      # Document storage
└── docker-compose.yml

---

Traficom's Challenge Solution

Our Solution: AI automation providing:

• Consistent evaluation standards
• Comprehensive compliance assessment
• Data-driven decision support
• Reduced inspector workload

---

License

This project is open-sourced under the Apache 2.0 License.

Primary design, backend architecture, compliance analysis engine, embeddings pipeline, and UI/UX concept were created by Tan Anh Nguyen during the Traficom Challenge at Junction 2025.

The core concept, system outline, and technical direction of CompliScan-AI originated from and were led by Tan Anh Nguyen.

See LICENSE for full legal terms.

Contributors

Frontend implementation:

• Luu Hoang Nhat Vi

Additional contributions in presentation and documentation by:

• Chambit Oh