AI Data Analyst
Customized Chatbot | 2024
Client: MylesAI
Skills Used: Python, OpenAI Assistant, Streamlit
The Problem
I wanted to develop a quicker and more efficient way to analyze data than writing code in Python or R, or manually creating functions in Excel. By creating an AI data analyst, I aimed to simplify and speed up data analysis, making it accessible to those without extensive coding knowledge. This approach allows users to quickly generate meaningful reports and visualizations, freeing up time to focus on making data-driven decisions. The goal was to empower everyone, regardless of technical background, to harness the power of their data through simply talking to the data.
The Solution
In order to properly train the POC chatbot, we needed data from both the ATL311 website and the Atlanta.gov website. We used a BeautifulSoup, LangChain, and ChromaDB to create a script to extract relevant information from entirety of both websites and store that information in an database for the chatbot to query.
​
After the database was built, we then used the OpenAI API to query the database. We also created guardrails for the chatbot so that it only answered questions related to Atlanta. The videos below demonstrate the chatbot's capabilities which include answering FAQs, providing sources for answers, drawing from multiple sources for answers, and having full conversation about Atlanta-related topics.