Key Points:
- Meta’s first-ever LlamaCon Hackathon drew 238 developers and innovators to San Francisco.
- Teams competed for $35,000 in prizes by building AI-powered projects using Llama 4 models.
- OrgLens, Compliance Wizards, and Llama CCTV Operator took top prizes, while Geo-ML won best Llama API usage.
Inside the AI Hackathon That Sparked Global Innovation
Meta’s inaugural LlamaCon Hackathon concluded with four standout teams claiming top honors after a 24-hour coding sprint that pushed the limits of generative AI. Held in San Francisco, the event convened 238 developers from across the globe, challenging them to build cutting-edge applications using Meta’s Llama 4 Scout, Maverick, and Llama API tools.
The hackathon followed LlamaCon, Meta’s new global developer conference focused on open-source artificial intelligence. Of more than 600 registrants, 44 project teams competed in the coding challenge, which offered $35,000 in cash prizes and industry-wide recognition.
According to Meta, participants had to deliver fully functional AI solutions leveraging the company’s Llama models within 24 hours. Judges from Meta and partner firms like Groq, Lambda, and SambaNova evaluated submissions based on innovation, functionality, and effective use of Llama APIs.
The Winners: Smart Matching, Fraud Detection, and AI Video Monitoring
1st Place – OrgLens:
OrgLens created an AI-enabled internal expert matching system. By analyzing organizational data—from Jira tickets to GitHub commits—it constructs digital profiles and lets users interact with digital “twins” of team members before connecting. The app was developed with React, Django, and the Llama API.
2nd Place – Compliance Wizards:
This team built an AI fraud detection platform that flags suspicious transactions and uses voice AI for user engagement. It integrates Llama API’s multimodal features for identifying clients’ potential links to criminal activity through uploaded documents and real-time news analysis.
3rd Place – Llama CCTV Operator:
Led by developer Agajan Torayev, this team used Llama 4’s multimodal vision features to automate security camera monitoring. Operators can define events in plain language, and the system detects relevant actions in surveillance footage without additional training.
Best Use of Llama API – Geo-ML:
Geologist William Davis won recognition for Geo-ML, a project that parses massive geological reports to generate 3D maps of mineral sites. Davis said he relied on Llama 4 Maverick’s long context capabilities to turn research into structured, visual data.
“Having an LLM that can do this work for me in the background will be really excellent,” Davis said.
As reported by Meta AI, the hackathon also highlighted ingenuity outside of the top prizes. Team Concierge brought their own GPUs to the event and fine-tuned a Llama 4 Maverick model using Meta’s recently released Synthetic Data Generation tool.
What’s Next
Developers can apply to join the next Llama Hackathon, scheduled for May 31–June 1, 2025, in New York City. With a growing ecosystem of open-source AI tools and support from global tech leaders, Meta’s Llama series continues to accelerate real-world generative AI applications.
(With reporting from Meta AI and Meta Developer Blog)
A global media for the latest news, entertainment, music fashion, and more.
