Restaurant menus can often be complex, filled with a variety of cuisine names, wine selections, ingredients, and detailed cooking descriptions, which can be overwhelming for customers. DineBot offers a solution by guiding customers through the menu, explaining dishes, making personalized recommendations based on their preferences, and assisting with placing orders. Whether deployed on the restaurant’s website for online orders or used on-site during times of staff shortages, DineBot can significantly enhance the dining experience by easing the decision-making process for customers. DineBot utilizes Retrieval Augmented Generation(RAG) to provide LLM with contents on the menu related to users’ query.
Nowadays people usually have a long list of emails to handle, which would be a tedious and time-consuming job to do, especially in a situation when people are asked to focus on a specific topic of their emails. Therefore, I developed this Emailcarer app which can rank the email lists dependent on its relevance to the user-selected topic. Based on the relevance score given by the IBM Granite LLM, the emailcarer will rank emails based on user-selected topic. It's developed as a google script add-on thus user can use it directly on their gmail workspace. I will publish this google script add-on later. Link will be given in the github repository.
Cleaning room could be a troublesome task. There are lots of items inside a used hotel room. Some items belong to the hotel, and some items belong to the customers that cleaning robot should leave. Some items are dirty items need to cleaned and refreshed while some items are garbage that just need to be thrown away. RobotSight uses Aria model to strengthen the visual analysis capability of a cleaning robot. With the comparison of used room and the room’s original condition, it can help cleaning robot to categorize the items inside the hotel room and create tailored cleaning schedule according to each item’s classification.