Generative AI Project - Google

Kasper Mechanical LLC recently completed training in Generative AI with Google Cloud products. It was an intensive 5 day course covering: Day 1 - Prompting, Day 2 - Classify Embeddings and Similarity Scores with Retrieval Augmented Generation, Day 3 - Function Calling with Gemini API, Day 4 - Fine Tuning a Custom Model, Google Search Grounding, Day 5 - Capstone project. This post is to cover the capstone project I created to put together an agent trained on my professional information. I will spare the fine details but the condensed description is that I used chromadb to store the information (mostly from my resume) to create a set of documents. I then used the embedding function and Google Gemini 2.0 to create a model that is able to recognize the text and respond to my questions based on the information I embedded. Possibly in the future, I'll get better at this and you will see a widget on my website with a helpful character to answer questions and accept project information. Check out some of the excerpts of code from my project below! Let me know if you have any thoughts, tips, or critiques!

This is where I fed it all of my data into documents:

It learns to regurgitate, but it's a very dry response.

So we tell it (rather Google tells it, I didn't come up with the prompt language myself, but I liked the effectiveness of this one in an example used in the course) - English please!!

And finally the model says, oh okay. I can answer the query in that prompt style.

Fascinating!!