I wrote about the process of producing the solutions in this repo here on my website.
This repository contains my proposed solution for the data science case at the second interview for the role as a data scientist at Laerdal Copenhagen.
The repo containing the case can be found at maximillian91/img-txt-categorisation-chat. Here you will also find the exercise descriptions to contextualize my solutions if interested.
My approach to this challenge was to explore solutions in notebooks and present my chosen solution in a Streamlit application. This resulted in a collection of notebooks in the notebooks-folder. Much of this is "quick and dirty"-work as the case had a fairly short time frame.
Ollama is used for the chat exercise. Run ollama serve followed by ollama run llama2:7b-chat in order to load and serve the model that is expected of the code. It uses Langchain and Streamlit Chat elements to glue everything together.