In this session, we'll discuss chat-based data science modeling and results interpretation. Data science modelling and experimentation often happen in an interactive notebook-based (REPL) environment. All of these steps involve writing small chunks of application code in the interactive environment. This work could be augmented by introducing a chat-based, interactive agent, which could interpret instructions, help with the hypothesis and answer related questions in plain English with a chat functionality. This agent embedded in the interactive environment can help with selecting features, interpreting results, providing consolidated answers related to technologies and technical questions and, most importantly, could generate code snippets to aid and speed up the data scientist's work. Once the modelling is done and the results are shown to business analysts and stakeholders, it is either up to the data scientist to try to convey the results in simple terms or feed them into a data warehouse for further interpretation. This process can be elevated by providing business users access to this interactive environment where they can interact with the results by asking questions. These cases are great examples of extending the capabilities of an analytics platform by allowing users to speed up modelling work and share the results directly with end users.
