Why every developer should be a data scientist
summary
Argues that designing systems around actual data yields concrete wins: Twitter switches to resize on the fly, keeping variants 20 days and cutting storage by ~4 TB/day and saving about $6M/year; a fraud-detector replaces heavyweight graph DBs with a union-find set, getting ~1–2 ms responses and tiny memory use; a team that ignored data wastes time by swapping to Elasticsearch; realistic traffic follows a Poisson process not uniform bursts; cockpit sizing finds essentially 0% true “average” pilots.
AI-written summary · about
chapters (11)
- 0:00Introduction and background
- 1:59Designing systems around data
- 4:28Distributed systems as programs
- 7:20Twitter's data-driven refactor
- 10:27Anti-pattern in data handling
- 13:25Revelant's fraud detection solution
- 18:11Why data science skills matter
- 21:06Cognitive biases in data analysis
- 24:15The myth of the average pilot
- 27:08Data science as a developer skill
- 31:48Key takeaways and resources
related talks
22mWhy is a good data scientist a package of professions?
19mA world of data
42mHow to be a 10x Data Scientist
32mThinking Like a Data Scientist
51mFrom Data Confusion to Data Intelligence
40mMistakes were made: Data science ten years in
23mBig data and big insights
43mFrom instinct to informed decision: launching data in your startup