Zero One Hackathon 🏆
Predicting insurance funnel drop-off in real-time and delivering personalized nudges to keep users converting — built for UNIQA at a 36-hour AI hackathon.
Role
Built in a team of 3 over 36 hours at the Zero One Hack Vienna hackathon.
Challenge
Needed to detect hesitation patterns from behavioral signals in real-time and trigger contextual nudges — explaining pricing, surfacing advisor availability — without latency killing the effect. User behavior had to be simulated at scale to test the system during the hackathon.
Outcome
Won 3rd place and €500. Combined Markov chain personas and LLM-driven personas via Ollama for rapid iteration, with a Next.js frontend and Python backend handling signal analysis and LLM reasoning.
Technologies used:
Next.js
Python
Ollama
