Value in Data Science Beyond Models in Production

on YouTube ↗23m

" summary

Argues that the durable value of data science is not pushing models into production but designing metrics, goals, experiments, prioritization and data products that steer $2M worth of engineering time toward real impact. Claims AutoML reduces the modeling moat and cites AutoML (H2O) is closing the gap and job counts (about 15,408 vs 17,931 jobs) as evidence. Recommends operationalizing logging, building R and RStudio Connect self‑service, treating data engineers as data product owners and owning experimentation and ethics.

AI-written summary · about

" chapters (15)

1. 0:00Intro
2. 2:07webmasters
3. 3:59where are they now
4. 4:43the death of webmasters
5. 6:05Google Panda
6. 7:44Data Science Hype
7. 10:17Data Engineers
8. 12:11Competitive Advantage
9. 13:13Data Scientist Responsibilities
10. 15:27Metrics
11. 17:42Selfservice tools
12. 19:41Ethics
13. 20:36The Future
14. 20:57superpowers
15. 21:18fist bumps

" related talks

• 45mData Science 2.0: From Outputs to OutcomesBill Schmarzo · SkillUp 2021
• 13mEnterprise-Level Data Science SuccessPrabhakar Thanikasalam · rstudio::conf(2022)
• 05mMaking data science useful.Cassie Kozyrkov · Strata Data Conference 2019 - London, United Kingdom
• 32mPanel: Make The Most Of Data ScienceEden Smith, Michael Terrell, Richard Pugh, Megan Stamper, Shorful Islam, Nicholas Deveney · Big Data LDN 2022
• 32mMLDS 2019: Pradeep Gulipalli of Tiger Analytics at Machine Learning Developers Summit 2019Pradeep Gulipalli · Machine Learning Developers Summit 2019
• 44mPeople, Technology, Culture: How to make Data Science workMario Meir-Huber · Data Science Conference 4.0
• 15mMachine Learning: 5 Additional Metrics to Avoid High‑Performing but Useless ModelsMarie-Anne Blette · Devoxx France 2023
• 36mAssuring Business Outcome Through AnalyticsPinky Sahu · Cypher 2017