The problem RapidRes solves

Booking travel is often a frustrating and time-consuming task, especially for elderly and disabled individuals. Navigating complex booking systems or managing multiple websites for trains, flights, and buses can be overwhelming. For those with mobility issues, these platforms can be difficult to use, often requiring precise clicks and form-filling.

The fragmented nature of travel booking, where separate sites are needed for different modes of transport, further complicates the process. This not only wastes time but also adds unnecessary stress to the experience.

RapidRes solves these issues by offering a simple, accessible, chat-based interface. With RapidRes, elderly and disabled users can effortlessly book trains, buses, and flights just by typing or speaking. The platform’s intuitive design ensures that users don’t need to struggle with complicated systems or multiple tabs.

Additionally, RapidRes supports advance reservations for even the earliest trips, offering peace of mind to those who need to plan well ahead. RapidRes is about making travel easy, convenient, and accessible—especially for those who need it the most.

Challenges we ran into

One significant challenge was the limited availability of APIs and datasets for essential services like food delivery and ticket booking. To overcome this, we engineered our own backend systems to replicate functionalities, listing train options, allowing us to develop and test the chatbot’s flow effectively.

Another hurdle was accurately capturing user context in conversation. We addressed this by implementing an intent detection system that identifies user goals based on keywords, which helped route conversations correctly. Additionally, we added context tracking to remember past interactions, allowing the chatbot to respond naturally and avoid redundant questions.

For real-time train details, we used Google’s Gemini APIs to retrieve relevant information, and for the chatbot’s conversational flow, we relied on Gemini models to understand and process user intents. However, we faced challenges in converting Gemini’s responses into structured JSON. We solved this by using parsing techniques and fallback prompts to refine the unstructured data into a usable format.

Finally, we implemented a fully automated captcha solver for the IRCTC API, enabling seamless interaction without manual intervention.