Kairos

what it is

AI-powered cultural event discovery platform for London, positioned as Spotify for live culture. A swipe-style taste quiz builds a personalised event feed with match scores, each backed by an AI-generated explanation of why the event fits.

what is unique

The edge is the discovery intelligence layer, not inventory or ticketing. A three-layer taste engine: an LLM enriches each event with vibe tags and social context, a multi-dimensional taste vector is built from the quiz, and a transparent scoring formula ranks contextually. Every recommendation comes with a generated reason, which makes it feel less like a listings site and more like a knowledgeable friend with taste.

how it was built

Next.js, TypeScript, Tailwind. Pinecone vector search over roughly 800 events, OpenAI text-embedding-3-large embeddings, Claude for the natural-language explanations. Event data enriched through an LLM pipeline. Deployed on Vercel.

gtm angle

This is the full stack end to end, not a feature bolted onto someone else's product: embeddings, vector retrieval, and an LLM explanation layer, shipped live. That depth is what lets a GTM person scope a roadmap and talk credibly with engineering rather than just market around the edges. The sharper GTM call sits in the ranking. The niche-event score multiplier, tuned so a big Ticketmaster listing does not drown out a more interesting small one, is the judgment every AI product forces: when to tune for business intent, not just raw accuracy.

from the live build

01 / 03

Eight questions, ninety seconds, then a feed of London nights built for one person.

02 / 03

The quiz opens on vibe rather than category, because how you want a night to feel predicts more than what genre it is filed under.

03 / 03

Every match carries a score and a written reason for it. The reason is the product: a bare percentage does not tell anyone whether to go.