Challenges

Challenge #1

DATES

3/7-4/19

FOCUS AREA

Desktop, Field, and Edge Applications

OVERVIEW

​In this challenge, your mission is to build innovative applications that harness Meta’s Llama Stack to run Meta’s Llama models locally on devices like desktops, field equipment, edge hardware, or tablets. These applications should be designed with local usage as a primacy function—think privacy, offline access, and fast performance. That said, it’s perfectly fine if they also connect to remote models, as long as they’re built to switch to local models when needed. The focus is on creating practical tools with the flexibility to go hybrid.

1. REGISTER2. DISCORD

Getting Started

1. Register Here
2. Join Discord
3. Submit your project by [4/19/2025] on Product Hunt

Any current undergraduate or graduate student -- or those who've recently dropped out -- can join. You may register solo or form a team of up to five members.

Project Ideas

These are just starting points - we're excited to see how you'll push the boundaries of what's possible with on-device AI. Your project could combine multiple approaches in mobile, wearable, or edge computing…or explore entirely new territories!

1. Smart Desktop

- Create a writing assistant that runs locally for offline editing and private drafts, but can switch to a remote model for extra features like style suggestions when connected.
- Build a desktop design tool that uses local models for quick rendering, with an option to tap remote models for heavier computations.

2. Privacy

- Develop a healthcare app that processes patient notes locally to keep data secure.
-  Design a finance tracker that runs locally for personal budget analysis.

3. Field and Forward-Deployed

- Create a tool for field technicians that uses local models for real-time equipment diagnostics offline, but can connect to remote models for detailed manuals or support when the internet is available.
- Build an agriculture app that runs locally to monitor crops on-site, with a remote option for weather data when connected.

4. Low-Resource

- Create an AI tutor for students that works locally on old desktops with no internet.
- Build infrastructure to help older phones or computers run compact Llama models locally.

4. Edge & IoT

- Create edge computing in factories that process real-time data where it’s collected. Switch to a larger hosted model only for complex evaluations or to aggregate data across sites.
- Privacy-focused sensors or assistants that run Llama models locally for tasks like voice recognition or anomaly detection, optionally reporting insights to a remote dashboard.

Why It Matters

Take ownership of a technological revolution that brings AI directly to billions of devices:

- Process data instantly, without server latency
- Work offline, anywhere in the world
- Keep sensitive information on users' devices
- Run smoothly on limited device resources

Resources & Support

Weekly Office Hours

- Join our live sessions via Discord or video calls for real-time Q&A with experts.
- Office hours to be announced on Discord.

Technical Documentation

- Access comprehensive guidelines for Llama Stack deployment
- View Github

Discord Community

- Chat with peers and mentors, share progress, and troubleshoot issues.
- Join Discord

Direct Inquiries

- Email us at llamachallenge@8vc.com for additional support.

Submission
Requirements

Timeline

- Participants can submit projects on three submission dates:  3/21, 4/11, 4/18.
- Submission dates are spaced out to ensure students have plenty of time to work on the challenge and submit multiple times.  

Submission

- Participants will take their code repository, record a demo video and create a product hunt submission before any of the three submission dates.
- We're leveraging Product Hunt so submissions are public and the community can engage with projects.
- Feel free to reach out to Product Hunt (link to challenge page) with questions.

Evaluation

- All submissions will be evaluated at the end of the challenge -- there is no priority for early submissions.
- We will notify the winner and two finalists a week after the challenge concludes -- see below for more details regarding the various prizes.

Prizes

1ST

- Finalist Retreat
- LlamaCon Tickets
- Mystery Prizes

2ND

- LlamaCon Tickets
- Mystery Prizes

3RD

- Mystery Prizes

Evaluation Criteria

We will evaluate all projects on the following criteria:

Technical Innovation

- Novel applications or improvements of Llama models
- Model optimization and efficiency
- Performance improvements
- Integration with hardware in innovative ways
- Quality of technical implementation

User Experience

- Intuitive and engaging interface designs
- Thoughtful handling of edge cases

Real-World Impact

- Solve meaningful problems
- Clear differentiation from existing solutions

FAQ

Intro