Overview
The Arm Create: AI Optimization Challenge is a Devpost hosted event inviting developers to build and submit projects that demonstrate optimization of AI for Arm powered platforms. The challenge is structured around three tracks that cover AI across real world, scalable cloud infrastructure, and on device edge experiences. The goal is to produce implementations that show measurable improvements in model size, quality, speed, and deployment efficiency on Arm powered hardware and ecosystems. The event is online, broad in scope, and aims to engage a global community of developers to explore Arm specific optimization opportunities.
Tracks
- Physical AI: Optimize AI for real world systems including robotics, embedded devices, sensors, simulation, autonomy, and edge environments.
- Cloud AI: Optimize AI for scalable infrastructure, including Arm64 cloud, inference performance, frameworks, agents, and production-ready developer workflows.
- Mobile AI: Optimize AI for on-device constraints, including performance, privacy, latency, battery efficiency, and local AI experiences on Arm powered phones, tablets, and laptops. Across all tracks, submissions should demonstrate clear optimization work with measurable improvements whenever possible.
Prizes and Evaluation
The challenge offers substantial cash prizes and recognition. Overall Winner, Runner Up, and category winners are awarded cash prizes, with projects featured in Arm community channels. Judging emphasizes Technological Implementation (40 points), User and Developer Experience (15 points), Potential Impact (20 points), and a WOW factor (25 points). Submissions should include a detailed description, a public code repository with a license, setup instructions for Arm powered environments, and, if available, a demonstration video.
How to participate
Developers must submit a project built with the required developer tools and meet the track requirements. A public code repository is required with clear documentation and licensing. Submissions should include a project overview describing purpose, features, and why it should win, along with setup steps to reproduce results on Arm devices. Track 1 and 2 require code access to source artifacts; Track 3 expects proof artifacts as described in the track requirements. Arm Performix is recommended for benchmarking references. The schedule spans from early June to mid August 2026, with a final deadline of August 14, 2026 at 4:00 PM PDT.
Schedule, access, and contact
Dates are published and can be viewed on the challenge page. The event is online, with resources and calendars available for participants to add deadlines to their personal calendars. For questions, participants are encouraged to contact the hackathon manager via email. The event is hosted by Devpost in collaboration with Arm, aiming to accelerate AI optimization on Arm-based platforms.
Who should attend
Developers, researchers, engineers, startups, and students interested in Arm optimized AI workflows, efficient models, and deployable AI solutions across embedded devices, cloud infrastructure, and mobile platforms. The challenge welcomes a global audience and encourages open source contributions and transparent demonstrations of optimization results.
