LLM developer Offline

Build a chatbot to make customer support more efficient. Start with a specific dataset and figure out which approaches work best.

First task:

Pick a public dataset such as Jira's "Jira-Software-questions" and experiment with different approaches to train a chatbot to answer questions based on the information in the dataset.

Approaches to try:

- in-context learning

- fine-tuning

- interaction between the retriever and LLM

Stack / libraries

- langchain, haystack, etc

- vector databases

- open source chat-bots like vicuna and koala.

- ChatGPT