​Siri and Alexa are boring. You can do better.

​LiquidMetal AI and SambaNova are hosting a 6-hour sprint to build voice agents that people actually want to use. Bring your wildest ideas for tool integrations - the weirder, the better.

​Event Schedule

​11:00 AM - Opening presentations from LiquidMetal AI and SambaNova (15 minutes)
11:15 AM - Building phase begins
3:00 PM - Submissions close
4:00 PM - Lightning demos from startups
4:30 PM - Hackathon demos and judging
5:00 PM - Winners announced

​Competition Format

​Participate solo or form a team. We'll award prizes in both categories, so choose the format that works best for you.

​What We're Looking For

​Build a voice agent that demonstrates:

• ​Natural conversation capabilities

• ​Smart context understanding

• ​Creative tool and API integrations (bonus points for unique implementations)

• ​Real-world problem solving

• ​Speed and responsiveness

​Who should join?

​Builders who want to ship a voice agent in 5 hours. Hackers ready to connect voice to any API imaginable. People who've built chatbots and want to make them speak. Developers who think voice is the next frontier.

​Join if you're ready to build the voice assistant you've always wanted - one that actually understands what you mean.

​Tech Stack

​Participants get free access to:

• ​SambaCloud: Low-latency inference powered by SN40L chips - get responses fast enough for natural conversation. Access DeepSeek, Llama, Qwen and other models with sub-second response times.

• ​LiquidMetal Raindrop MCP: Build directly in Claude Code with our MCP integration - write, test, and deploy your voice agent without leaving the AI assistant. Deploy to global edge locations for minimal latency. SmartMemory maintains conversation context, while serverless compute scales instantly.

• ​Hume Empathic Voice AI: Real-time speech that carries tone, cadence, and emotion. Powered by EVI 3 and Octave TTS, it interprets and responds with sub-second latency so conversations feel natural. Decode emotional signals across voice and expressions, then generate replies that adapt on the fly.

​Voice agents need speed. These platforms deliver the millisecond-level performance required for fluid conversations.

​Prizes

​Winners receive:

• ​WWE Championship Belt (because you're the champion of voice agents)

• ​$5,000 in Raindrop platform credits

• ​Direct access to both companies' engineering teams

​Getting Started

​Teams will receive API keys and documentation at check-in. You will also need a subscription to either Claude Code or Gemini to use the Raindrop MCP.

​Come build something people will actually want to talk to.