- The Next Project started with me meeting a senior dev at LAITS who focuses on AI projects.
- This was a good way of learning how to Integrate my Masters in Artificial Intelligence into my development work at LAITS
- The Goal is to develop a chatbot for the COLA helpdesk that can help assist in the process of staff having to manually sift through ticket requests and emails for basic help questions
- I am a user of Claude for some of my personal projects (https://abdurrahmaan-baghdadi.github.io/)
- Claude had downtime when I started to work on this project and so I decided it would be better to work on self hosting a model.
- From here I looked into the Qwen 3.8 ollama 27B model to host on my local machine
- this would not only continue progress in the project at a nice rate but also would avoid downtime problems caused by relying on claude
Below are some of the integrations that the my local LLm could support:
- Later along the line, Claude’s availability returned
- I decided to take advantage of having 2 LLMs at hand
- The goal was to have my claude application manage my Qwen interactions for me – trying to optimize the model and setup for my system but also train the model to handle the creation of a help page router (the actual machine learning algorithm that would be used to help users navigate to necessary pages)
- I iterated through various ideas trying different solutions
- Along the way I wanted to benchmark my approach
