 Hello everyone, my name is Livia. Welcome to this elevator pitch on customize your personal data science assistant bot. I developed this project together with my co-worker Oliver, we're two creative data scientists from statworks. Statworks is a consulting and development company for data science machine learning and AI based in Frankfurt and CERC. And together we want to make AI accessible to everyone and that's also why we created this data science chatbot. So how does this look like? Here you see a screenshot of this data science assistant bot. So essentially it's an assistant bot that helps you regarding any data science topics. Actually it's a shiny interface we created for GPT-3 to simply and quickly customize this data science assistant bot. So underlying is GPT-3 and we just use shiny to create this interface where you can easily adapt your bot. So we have certain features. First, let's have a look at the customization. So as I said, you can customize this assistant bot to your needs. So you can set different parameters. You have this column on the left hand side so you can, for example, choose the field of interest. For example, here machine learning is selected or you can choose which programming language you want to talk about. You can choose the answer length of the bot. You can also choose how friendly the assistant should be and you can select your knowledge level. So no coding is required. You can customize this assistant bot only by point and click. And during the conversation you also can easily change these parameters again and change the assistant on the fly. The chat itself is very easy. You have this box where you can enter your question on the left hand side and then you immediately get the answer in the chat window, which is on the right hand side. And there you can also see the whole history, the whole chat history. As I said, underlying this bot is the GPT-3 model, which is another aggressive language model developed by OpenAI in 2020. Until today, there is only a test version for selected users available, but soon I think they will release it for everyone. This model is really all about generating human-like text. And you also can use this language model for other use cases so it can also be used for text summarization or Q&A. And this NLP model is accessible via an API, which we also used for our chatbot here. So how this works is the following. We are using the text-in-text-out interface of this API. So every time the user enters a new question, a new API request is sent to the open API and then it sends a response with the answer of the chatbot. So to generate this answer, not only the new question is sent but also the chatbot description, which is made via the parameters you set for the chatbot and also the whole chat history. So these three things together are sent as a text-in and then as a text-out, you get the answer. So how did we implement this with Shiny? So we chose Shiny because it makes this very easy to create this interactive customization of this data science assistant bot. So you can really choose those parameters by click and point. So actually how this customization is done is that this description of a bot is a long string as you can see in the picture. So there you have a description of an assistant bot and there within you have different inputs and that's how you change this description. It's done by Shiny inputs. There you change these different parameters in the string within. You can also see an overview of the packages that we used and our backend is mostly inspired by Tesla Tech's Shiny chat. So there it also stores the whole chat history, which is very convenient because we need to send that also to the models API. So that was a really short presentation and I'm very happy to answer your questions if you have any. Thank you very much.