The Real-Time AI Chat Interface is an intelligent, multilingual transit assistant designed to bridge the accessibility gap in Maharashtra's public bus network. Built specifically to handle regional challenges, the platform allows commuters to type natural queries in English, Hindi, and Marathi (fully supporting native Devanagari script).

Behind the scenes, a hybrid NLP engine instantly processes user intent and utilizes Nominatim and OSRM to provide end-to-end, multi-leg journey planning that connects remote villages to major city transit hubs. Beyond route lookup, the system calculates dynamic, accurate transit fares and integrates ticket booking. Once a booking is finalized, the system automatically triggers real-time delivery of travel itineraries and e-tickets directly to the passenger's mobile device via WhatsApp and SMS.

For this hackathon, we are upgrading our legacy rule-based and fuzzy-matching NLP infrastructure to advanced open-source Large Language Models (LLMs) like DeepSeek-R1 and Llama 3 using Crusoe’s Managed Inference. This integration will significantly improve the accuracy of multi-turn conversation tracking and complex Named Entity Recognition (NER) for multi-lingual village-to-city route queries, creating a truly inclusive, hands-free transit solution.

Built With

  • .net
  • and-marathi-to-provide-automated-door-to-door-journey-planning-(connecting-remote-villages-to-cities)
  • and-real-time-ticket-delivery-via-whatsapp-and-sms.-during-this-hackathon
  • c#
  • css3
  • dynamic-fare-calculation
  • hindi
  • html5
  • i-hope-to-learn-how-to-seamlessly-implement-and-optimize-managed-open-source-large-language-models-(llms)-like-deepseek-r1-and-llama-3-via-crusoe's-infrastructure.-my-specific-goal-is-to-upgrade-our-legacy-rule-based-nlp-parser-into-a-highly-accurate
  • javascript
  • multilingual-transit-assistant-that-simplifies-public-bus-travel-across-regional-and-rural-networks.-the-platform-processes-natural-language-queries-in-english
  • python
  • xml
Share this project:

Updates