Financial Filings Analysis

This repository contains a minimal, end‑to‑end pipeline to analyze sections in 10‑K filings for several retail companies over the past several years.

What this gives you

• Scripts to download 10‑K filings from SEC EDGAR, extract section intended, and save to disk
• LLM-based tagging via Anthropic Claude
• EDA helpers and plotting stubs
• A CLI (run.py) to orchestrate the pipeline
• A standalone script (run_report.py) to generate a report from existing data

Data and Outputs

• Processed data that contains extracted section is stored in data/processed/ as JSONL files
• Final results are under outputs/, where plots and tables for individual companies and aggregated evolution summaries are stored

Project Layout

aibizops/
├── data/
│   └── processed/          # JSONL with extracted section per company-year
├── notebooks/
├── outputs/
│   ├── figures/            # charts
│   └── tables/             # CSVs/TSVs
├── src/aibizops/
│   ├── __init__.py
│   ├── edgar.py            # download & locate filings
│   ├── parse.py            # extract specific section
│   ├── llm.py              # Claude classification
│   ├── pipeline.py         # high-level pipeline steps
│   ├── eda/
│   │   ├── plots.py
│   │   ├── novelty.py
│   │   └── stats.py
│   └── utils/
│       ├── io.py
│       └── text.py
├── requirements.txt
└── run.py
└── run_report.py

Notes & Assumptions

• We focus on annual 10-K filings and only certain sections.
• We match multiple years by walking EDGAR submissions and selecting the last several 10-Ks.