SixtSense

Inspiration

We all believe that AI can simplify people's lives. SixtSense is the embodiment of this vision. In fact, it integrates an LLM wrapped in a langchain system to recognize and best meet the needs of a user about to rent a vehicle, also pointing out features and safety measures that they would never have known about otherwise. Our goal was to simulate the activity of a salesperson: in fact, just like the agent at the front desk, our model wants to get to know the customer better in order to offer them the best alternatives based on their interests and personal characteristics.

What it does

Our solution is a pre‑check‑in application where customers log in with their booking ID a few hours before pickup and interact with an AI sales chatbot, built by wrapping an LLM with langchain, that uses customer and booking data to recommend the best vehicle, protection package, and add‑ons. To rank vehicle suggestions, the model follows a hybrid approach: in fact, we combined OpenAI API calls to a deterministic scoring function. As the conversation continues, the chat-bot adapts to each customer, collects new data in real time, recalculates scores, and updates recommendations to highlight the most relevant up-sell opportunities.

How we built it

We developed the back-end using Python as the programming language and libraries such as langchain to implement an AI model using OpenAI APIs and NumPy to streamline calculations. We connected the “skeleton” of our program to the graphical interface using DRF and Django, with which we managed our API and URL requests, and we combined MVC and MCP technologies to make an agentic application. As for the front-end, we created a modern web app using TypeScript, React, Vite and TailwindCSS.

Challenges we ran into

The path to developing our solution was not linear at all. The first problem we faced was writing the basic prompt for the LLM we decided to integrate: finding the perfect combination of instructions was a very delicate task. In addition, we underestimated the connection between the back-end and front-end technology: this necessary task took us a long time, mainly due to technical difficulties related to the validation of our API.

Accomplishments that we're proud of

We are very proud to have implemented not only the core functions we promised ourselves, but also some features to improve the efficiency of the system and the overall user experience. We are very excited to have created a service that, thanks to its modularity, can be implemented in sales agents, but also in other areas of the market. Although it initially seemed like an almost unattainable goal, we can be satisfied with the integration of an AI Agent via langchain, because it proved to be the key to our project.

What we learned

We are very happy with how well we worked as a team, even though it was the first time for all of us. It was easier than expected to divide up the tasks, and we achieved excellent team harmony. Each of us, whether working on the front-end or back-end, was able to improve our existing knowledge and learn new things, especially regarding the scalability and integration of a full-stack application. Still on the technical side, we gained a deep understanding of langchain technology and learned how it is an almost necessary resource for solving challenges of this kind.

What's next for SixtSense

Our hope is that SIXT will integrate our technology into its vehicle rental services to create a unique and unparalleled user experience. Then, data collection is crucial for future developments: the more high-quality data we can obtain from users, the better and more personalized the service will be. To enhance this aspect, it would be interesting to exchange relevant data between companies operating in the travel and alternative transport sectors.

Built With

• django
• javascript
• langchain
• python
• react
• typescript
• vite