Overview

Before chatbots were an API call, building one meant understanding the whole stack — intents, training data, response logic. ChatbotV9 is that learning ground: a notebook-first environment for prototyping conversational models with Python NLP libraries, designed so you can change a dataset or a model and see the effect in the same breath.

Tech Stack

Challenges

• Designing conversation flows that feel natural instead of scripted.
• Wiring machine-learning models into an interactive notebook workflow.
• Cleaning and preparing conversational datasets that actually train well.
• Handling edge cases and off-topic inputs without the bot falling apart.

Solution

The project uses Jupyter for tight, iterative loops — tweak, run, observe — with standard Python NLP and ML libraries to build and train the models. A modular structure keeps data, model, and interface separate, so swapping datasets or trying a different approach never means rewriting everything.

Outcome

ChatbotV9 made experimenting with conversational AI genuinely fast, and it's the project that taught me the fundamentals I still lean on when grounding modern LLMs — what good training data looks like, why intent boundaries matter, and exactly where naive bots break.

What I'd do differently

This was pre-LLM thinking, and it shows. Knowing what I do now, I'd rebuild it around a retrieval-augmented LLM with a small, well-curated knowledge base and a real eval harness — which is, almost exactly, the architecture I later used for the AI assistant on this very site.