Back to projects

TripMate

Travel quote automation built in 48 hours at SingHacks 2025

TesseractOCR · Groq · OpenAI Whisper · ElevenLabs · Python

Overview

TripMate automates the tedious process of getting travel insurance quotes, reducing what typically takes hours of form-filling to under 2 minutes. Built during SingHacks 2025, the system extracts information from travel documents and interacts conversationally with users to generate accurate quotes.

Technical Highlights

  • Document Intelligence: Implemented TesseractOCR to extract passport details, flight itineraries, and other travel documents, automatically populating quote forms.

  • Voice-Enabled RAG Pipeline: Combined OpenAI Whisper for speech-to-text, a retrieval-augmented generation system for policy information, and ElevenLabs for natural voice responses.

  • Real-Time Policy Retrieval: Built a system that queries multiple insurance providers' policies to find the best coverage options based on user requirements.

Achievement

SingHacks 2025 - 2nd Runner-Up

Technical Challenge

The Problem: Insurance quotes expire, but conversational flows are unpredictable. Users pause, change details mid-flow, or need multiple attempts at payment. Each delay risks an expired quote and a frustrated user restart.

The Solution: We built a 3-tier pricing model derived from live Ancileo API data: Standard (Elite ÷ 1.8), Elite (base), and Premier (Elite × 1.39). Fresh quotes are generated frequently to avoid expiration, and Stripe webhook confirmation triggers automatic policy creation — no manual post-purchase workflow. The pgvector-powered RAG system lets users query 72K+ historical claims for coverage guidance.

Why It Matters: Separating pricing calculation from quote fetching meant users could browse options without race conditions against expiration timers.

Gallery

TripMate at SingHacks 2025TripMate Demo