Problem
Automation risk and actual job loss get conflated. The question: who is really most exposed when AI takes over, and how has that map shifted from older automation to generative AI?
Approach
Merge three authoritative datasets (Frey & Osborne automation scores, BLS 2024–2034 employment projections, ILO 2025 GenAI exposure index), engineer exposure features, then cluster and regress to separate risk from realized loss.
What I built
- A fetch → clean → analyze → mine pipeline producing 14 figures (including then/now comparisons).
- K-Means clustering and linear regression over the merged dataset.
- A 6-tab Streamlit dashboard and a Quarto report (HTML + PDF), with an annual GitHub Actions auto-refresh.
Result
Central finding: automation risk and job loss are related but not the same. The most-threatened workers are the least equipped to adapt, and GenAI has shifted exposure onto knowledge, analyst, and creative roles once thought safe.