Inspiration
Our team is passionate about AI and agents enhancing data connectivity and accuracy within ONE Record, and this hackathon was the perfect opportunity to showcase our skills. Having worked closely with ground handlers, we know that ULD damage assessment is a significant challenge, and we wanted to create a tool to help field operators. Additionally, the decentralized identity challenge intrigued us due to our prior experience with OpenID and our desire to tackle a steep technical problem.
Our inspiration came from real-world experiences, including warehouse visits, past work with computer vision for ULD damage assessment, and discussions with ULD Care and Jettainer about their operational challenges. The digital identity challenge allowed us to explore innovative solutions for distributed identities without relying on federated systems or extensive SSO integrations.
What it does
LUNA is an AI-powered assistant designed to guide users in assessing ULD damage efficiently and accurately. By leveraging voice, text, and image inputs, it reduces the risk of unnecessary repairs and improves operational efficiency. Users particularly appreciated its voice capabilities and its ability to process images, such as reading ULD numbers or analyzing damage.
On the digital identity side, our solution enables participants to share Verifiable Presentations from Decentralized Identifiers (DIDs). This eliminates the need for direct connections between actors in the ONE Record network, enhancing security and trust without requiring centralized user databases.
The entire system is accessible via a user-friendly interface that works on both desktop and mobile devices. With just a button press, users can chat with the agent, upload photos, and get guided step-by-step by the AI.
How we built it
We used Python as the primary language for development. For LUNA’s AI capabilities, we integrated GPT-4 for natural language processing, Whisper for speech-to-text functionality, Chainlit for the user interface, and LangGraph for agent orchestration. We also explored Kokoro for text-to-voice responses. For the decentralized identity solution, we implemented a proof-of-concept mimicking OpenID Connect using JWTs with FastAPI after experimenting with Keycloak and Python Dex.
To integrate with ONE Record, we utilized a local Docker Compose implementation of its open-source framework. We adapted it for cloud deployment and connected it via NE:ONE Play UI and APIs. However, this required patching LO on ONE Record to enable seamless functionality.
Challenges we ran into
One major challenge was gathering comprehensive documentation on ULD serviceability criteria to ensure LUNA could guide field operators effectively. This required extensive research and interviews with experts. Additionally, understanding how to patch ULD statuses in ONE Record proved complex.
For the decentralized identity solution, grasping the concepts of DIDs and Verifiable Presentations was difficult due to limited implementations and evolving standards. We spent significant time setting up libraries and experimenting before successfully implementing a proof-of-concept.
Time constraints were another hurdle. To overcome this, we set strict priorities with bonus objectives and worked tirelessly—sometimes overnight—to meet our goals.
Accomplishments that we're proud of
We are proud of creating a solution that resonates with end-users. Early feedback from customers highlighted LUNA’s accuracy in assessments, helpful guidance, and voice capabilities as standout features. For example:
- Users praised its ability to process photos of cargo net tags accurately.
- The assistant provided clear guidance on serviceability checks.
- It impressed users by reading dates from photos and responding with actionable insights.
We are also proud of tackling two challenges within the hackathon timeframe by splitting our team effectively. Despite limited time, we delivered both solutions successfully while maintaining quality.
What we learned
This project reinforced how quickly AI solutions can be implemented to deliver tangible benefits in real-world scenarios. LUNA demonstrated how AI can empower—not replace—field operators by making their work faster and more efficient.
We also learned about the potential of Decentralized Identity and Verifiable Presentations as mechanisms to connect identities across industries without requiring direct connectivity or federated systems.
Finally, this experience highlighted the importance of communication and teamwork. By respecting each team member’s strengths and talents, we collaborated effectively despite never having worked together on such an intensive project before.
What’s next for LUNA
Our next steps include enhancing LUNA’s capabilities by adding voice responses for users who prefer audio feedback. We also plan to expand its functionality to include reporting capabilities for damaged goods or dangerous goods checklists—making it an even more versatile tool for field operators.
On the digital identity side, we will participate in IATA workgroups to share our insights from this hackathon and contribute to shaping future standards.
Finally, we are open-sourcing LUNA so that the entire air cargo community can benefit from it. We welcome collaboration with industry stakeholders interested in further developing these capabilities.
Test it out!
You can test the project using the following link: [https://dublin-hackathon1.northeurope.cloudapp.azure.com/login] Ignore the certificate warning. To authenticate, use the login:bob password:1234
Log in or sign up for Devpost to join the conversation.