Hello everyone,

Thank you for registering for the Melange On-Device AI Hackathon. We’re excited to see what you build with on-device AI.

Below are the final instructions, submission requirements, and timeline for the hackathon.

What You Should Build

Participants are expected to build an AI application that runs directly on a mobile device using Melange.

Your project should demonstrate on-device AI inference, meaning the AI model runs locally on the device rather than relying on cloud inference.

You may:

• Use models from the Melange public model library
• Upload and deploy your own custom model
• Build any type of AI application as long as the AI inference runs on-device

Examples include:

• On-device AI assistants
• Privacy-preserving AI tools
• Translation or summarization apps
• Computer vision or speech applications
• Safety layers for LLM apps
• Any creative on-device AI use case

You do not need to build a full production application. A working prototype or proof-of-concept is perfectly acceptable.

• Judging Criteria
  Projects will be evaluated based on: • Use of On-Device AI — how effectively Melange is used • Technical implementation — quality of the prototype • Creativity and idea — novelty of the use case • Demo clarity — how clearly the project is demonstrated