In this article, I would like to introduce my new AI Bot, which is an agent AI that I have built almost from scratch. Obviously, the model is trained for general purposes; however, I have introduced augmented generation (also known as RAG) to improve the accuracy and to enhance the responses with more factual information.

This can answer questions about my blog. It can answer questions about my past work experience. Also, it’s not limited to that, but it can answer questions based on the training data that we supplied to the model. I can change the model at any time in the future depending on the cost and time available, and depending on performance and accuracy for future models.

I believe this model and the architecture is very secure and scalable since it has been developed by me with a security-first approach. It demonstrates the ability to build AI agents from scratch, since this bot doesn’t use APIs like ChatGPT or Claude. To be very transparent, this website might experience slower performance next to big tech, but the idea here is not to compete with those models but to provide a simple yet useful personal project to demonstrate capabilities. See the image below for the system’s architecture.