The Hackathon was a 2-day event, although we made this in probably one day for the purpose of experiential learning. This was my first time making a Python project for a real-world use case, so it was valuable in its own right!
For the AEP challenge, then, our objective was to take in a 20,000-line CSV data source, parse it, then categorize each event listed based on its safety risk by the presence of "high energy" situations, using the power of LLMs. These events had lengthy descriptions and titles, with many being red-herrings. We used Ollama with llama3.2:3b running locally to maintain data privacy. Unfortunately, I can't share the source data CSV at American Electric Power's request for privacy. However, I am happy to share more about the process.
This was run on Python 3.12 (although the version shouldn't matter a great deal), with ollama, csv, pandas, and plotly.express libraries installed and imported.
Please see chat.py to see the main processing and llama3.2 prompting script and graph.py to see my graphing methodology.