Using Generative AI to Increase the Impact of Your Data Science Work
summary
Uses Gemini (Gemini 1.5 Flash for high-throughput calls, Gemini Pro for deep reasoning) plus Vertex AI to read plots and generate plain-English explanations of model outputs: XGBoost predictions of Olympic medal counts with Shapley-value attributions. Produces 206 country reports in around seven minutes, outputs in ~40 languages, and flags over- and under-performers (e.g., Japan −7.7 medals vs. expectation). Provides runnable R/Python snippets, cloud PNG input, and examples of scaling single-plot explanations to a global summary.
AI-written summary · about
chapters (12)
- 0:00Generative AI for data scientists
- 1:25AI in data science communication
- 2:09Google's Gemini features
- 3:38Gemini for podcast prep
- 5:28Olympic medals prediction
- 7:57Explaining with Shapley values
- 8:19Stakeholder-friendly explanations
- 10:26Translating explanations
- 11:08Scaling to all countries
- 13:53Post-Olympics AI analysis
- 15:41Key stories from 206 plots
- 18:50Summary and next steps
related talks
18mIntroduction to Gemini APIs and AI Studio
01h 00mHow to work with generative AI in JavaScript
23mScaling GenAI Applications and Agentic AI for Business Competitive Differentiation
01h 45mAI Engineering with the Google Gemini 2.5 Model Family
11mapproach to generative AI
21mHow Google DeepMind is researching the next Frontier of AI for Gemini
17mUnlock the power of Generative AI: From understanding its principles to utilizing its power
15mBuilding with the Gemini API and AI Studio