Last week, we tried a simple but intense hackathon format with the AI Engineer London community: A lightning fast voice AI hackathon. No warm up, presentations, or slides. Just 90 minutes to build something real.

60 developers arrived at Cloudflare’s HQ in London, formed teams, and immediately started building. Judges were kept on their toes, circulating the room and watching demos as they came together. The vibes were chaotic, collaborative, and a ton of fun.

By the end of the evening, we saw over a dozen voice AI apps and awarded prizes to eight teams.

Here’s what the community built.

Overall winner: EchoDiary by Anshul Yadav

EchoDiary took home the top prize with a real-time, voice-first emotional journaling companion. Rather than typing out feelings, users speak and the agent responds with empathy and understanding.

Runner-Up: OmniForm by Nathan Amin

OmniForm secured runner-up position with a voice agent that fills out web forms for you, live. The demo showed the agent navigating complex multi-step forms entirely through conversation.

Best Use of Cloudflare: Eliza Redux by Anup Jadhav

Inference Forge demonstrated exceptional use of Cloudflare's infrastructure to power their voice AI application.

Eliza Redux is a real-time voice agent that provides crisis support through natural conversation, grounding techniques, safety assessments, and emergency service location.

Visit Anup's blog to watch a demo video and read a detailed technical breakdown of the project.

Best Use of Deepgram: StoryTeller by Catherine Tranfield and Prakash Verma

StoryTeller won for their implementation of Deepgram's speech-to-text capabilities, demonstrating the power of accurate transcription in creating engaging narrative experiences.

Congratulations also to Fire fallback, Riley, Eddodai, and Big Ben London for their creative projects and strong performances at the hackathon.

Voice + UI creativity

Beyond the winners, all teams brought a lot of creativity to their multi-modal demos exploring really useful and fun applications of voice AI:

• An app that explains art through conversation

• A crisis support line accessible via voice

• A fun quiz game that plays like a conversation

• An AI sales coach that role-plays scenarios

• A voice escape room experience

• Voice agents designed specifically for dog owners

What we learned

This was our first lightning hackathon, and the format proved its worth. The compressed timeline and goal of getting to real demos kept everyone focused on building.

It was also fun to see what people can build when they don’t need to worry about infrastructure (or food and drink): Teams used Layercode CLI to get up and running quickly, helping focus solely on building their core product experience — no one spent time configuring audio pipelines or debugging latency issues — they just built products.

Every team used Layercode to able to power their voice agents, connected their custom logic via webhook, and immediately started building their core experience.

Thank you!

Massive thanks to the AI Engineer Meetup London community, and to our partners and sponsors Rime, Deepgram, and Twilio for supporting the event.

Also, special appreciation to our judges who stayed on their feet for the full 90 minutes: Anna Cachadina Abello, Anni Chen, Shreman Shrestha, Confidence Okoghenun, and Jan Siml. Your feedback and energy made the event.

Finally, huge thanks to Thomas Ankcorn for hosting us at Cloudflare HQ and Micah Smurthwaite for helping coordinate everything behind the scenes.