VoiceDine

Inspiration

We've all spent time bickering in a group when planning a holiday with no clue about the local food scene and specific dietary requirements you need to meet. It's safe to say that's a headache that no one wants to deal with. Our project eliminates it - in a human first solution!

What it does

Take your conversations and use them to give you the right information so you can actually make a fair decision that meets everyone's requirements. Going even further - once you've made a decision, have the booking AI make the phone call and reserve a table for you.

How we built it

A lot of vibe coding - a simple tech stack which interacts with more complex external APIs such as elevenlabs, exa.ai, twillio. We collaborated well using github, version control and abstracted the problem into parallelisable parts - and traded when things got hard. AI was a huge focus for us here as it drives a lot the value, both in voice and LLMs.

Challenges we ran into

Speed of response - LLM APIs took a while to return a response whereas we wanted it to be real-time
Transcribing multiple speakers using ElevenLabs in a crowded environment!

Accomplishments that we're proud of

A strong, sleek interface
Voice ai integration - agentic reservation booking

What we learned

Team management was SUPER important... had a team member leave at the start too which changed the dynamic but we adjusted fast
Elevenlabs also has surprisingly a lot of models so picking between them was interesting.
We used both ai tools, and used many api calls here, so especially important was reading through documentation on SOTA tools which LLMs will hallucinate on.