Omar Azab

Making an LLM Based App That Helps Users Repair Bikes

Started Dec. 11, 2024

Abstract or project description

Bikes are vehicles with many moving parts that often need tuneups and repairs. Many bikers have to make constant, costly trips to the bike shop to repair their bikes, and it's often for problems that they can fix themselves, if only they knew how to. This Is why I am making a bicycle repair app that uses LLMs such as Deepseek or chatGPT to help users diagnose and fix problems on their bike. LLMs are already capable, and have large available databases, so using them to help fix bikes is the perfect option for my app. In the app, the user will send an image of the part of the bike with issues. The AI will then analyze the images, and ask follow up questions based on the problems the user and the image are describing. Then, it will provide a detailed summary on what is wrong and all the options to to fix it. Some challenges with making this app is that currently, most LLMs are more text based, and for this app, both text and image processing are needed. Another challenge is providing a structured method for working with the LLMs directly through the app, and having the LLM use the same process and format the questions the same way every time.