Join Akash Network and Venice AI for a one-day hackathon focused on building the next generation of decentralized AI agents and open-source systems.

This hackathon is for developers, researchers, and builders who believe AI should be open, composable, and permissionless — not locked behind centralized APIs.

Participants will build across two core tracks, both centered on decentralized AI agent infrastructure and open systems:

Akash Agent Track: 

• Create an autonomous or semi-autonomous AI agent deployed on Akash that continuously operates (not a one-off script) and is capable of task execution, monitoring, orchestration, or decision-making.
  
  
• The agent must take verifiable real-world actions such as authenticated API calls, multi-step tool usage, job scheduling, stateful resource management, or interactive user workflows — beyond simple prompt/response behavior.
  
  
• Demonstrate persistent state, fault tolerance, and compute independence by running the agent entirely on Akash Infrastructure (no fallback to centralized cloud compute), including redeploys or restarts without loss of core functionality.

   

  Outcome: Must be a working agent hosted on Akash Console and use a hosted model via the akashml API while meeting the criteria above.

Venice Track 

Private Multimodal Intelligence

Vision, Voice, and Reasoning Under User Control

Real intelligence is multimodal. Modern AI systems combine language, vision, speech, and video to understand and act in the world. Most existing multimodal pipelines centralize this data in third-party cloud systems, creating long-lived records of highly sensitive user information.

This track focuses on building multimodal AI systems that deliver advanced capabilities while preserving user control over data and execution.

Create applications that combine language, vision, speech, and video into coherent workflows while minimizing external data exposure and long-term retention.

Projects should demonstrate how multimodal intelligence can operate with:

• Controlled data flows

• Minimal retention

• Transparent processing

• User-owned storage

Projects may include:

• Voice-first agents with private visual perception

• Secure meeting and call assistants

• Local-first or hybrid perception pipelines

• Confidential video understanding systems

• Privacy-preserving creative studios

• Multimodal copilots for sensitive workflows
  
  

• An AI assistant that summarizes meetings without storing recordings

• A voice-controlled design studio with encrypted asset storage

• A video analysis system that avoids centralized raw footage storage

• Idea 20%
  Does the solution have the potential to solve a meaningful problem or demonstrate real-world value?
• Technical implementation - 20%
  How well was the solution implemented?
• Tool Use 20%
  Did the solution effectively use the sponsor tools?
• Presentation 20%
  Demonstration of the solution in 3 minutes
• Autonomy 20%
  How well does the agent act on real-time data without manual intervention?