Lightning Talks (02) & Closing Notes
summary
Quick scan: a live demo claims Feinman instrumentation monkey‑patches functions and draws execution in the browser, showing thousands of modules loading and JIT compile times; a paper claim that a two‑model confidence gating setup cuts large‑model compute ~33%; an anecdote where GPT‑3 few‑shot hits ~61% versus a trained 63% baseline; a practical Streamlit + ngrok demo exposing a local multi‑page app; a clear explainer of index‑aligned pandas rules that justifies groupby/transform/app behavior — watches that show code, numbers, and working tools.
AI-written summary · about
chapters (12)
- 0:00Lightning talks kickoff
- 1:08Instrumenting Python apps with Feinman
- 5:00Data science for tax optimization
- 11:01Netflix imports and procrastination
- 12:35Lightweight neural networks
- 16:16Open source with Data Umbrella
- 22:15Anomaly detection in business
- 28:36Learning Python with Khan Academy
- 31:08Will GPT-3 replace data scientists?
- 35:40Live-coded Streamlit app
- 42:18Pandas inconsistencies
- 56:29Closing notes
